Results

Results of Literature Search

Older Versus Newer Antiepileptic Drug Evaluation

A summary of search results is presented in Figure 2.

Narrative: Shows the flow of study identification and selection. The initial database search resulted in 4170 records from MEDLINE, 528 records from CENTRAL were interim reports of original trials included in the evaluation. Sixty-eight articles were included in qualitative synthesis. Of these, 49 trials were eligible for quantitative synthesis.and 1665 records from WEB OF SCIENCE. An additional seven records were identified from other sources. After duplicates were removed, there were 5733 unique citations eligible for title and abstract screening. The first phase of screening excluded 5505 records for the following reasons: 1565 records were not reports of new discovery in humans, 228 records were not in epilepsy patient population, and another 3698 records did not compare older to newer AEDs, and 14 reports did not report outcomes of interest. This process left 268 records to assess for eligibility by screening the full-text articles. The second phase of screening excluded 200 articles because 4 articles were not reports of new discovery in humans, 8 articles were not in epilepsy patient population, 19 articles were not comparing older to newer AEDs, 7 articles were not published in the English language, 101 articles were not evaluating outcomes of interest, 58 abstracts were excluded because no response was received from the authors, 1 article was a duplicate of an original trial included in the final evaluation and 2 articles

Figure 2

PRISMA flow diagram of citations identified and evaluated in the older versus newer search. AED = antiepileptic drug; PRISMA = preferred reporting items for systematic reviews and meta-analyses

Upon conducting the literature search to identify articles that compared older antiepileptic medications to the newer ones, we retrieved 5,773 unique citations, and another 7 citations were identified from other sources. During the title and abstract review, 5,505 articles were excluded, and 200 articles were excluded during the full text review. A list of articles excluded during full text review can be found in Appendix C. A total of 68 articles were found to match our inclusion criteria. Although no language restrictions were imposed in the literature search, seven articles were excluded from the full-text review because they were not published in the English language. Given the small number of articles excluded, their small sample sizes, and the traditional medicine versus complementary medicine topic, we do not anticipate that their inclusion would have impacted our results in a meaningful way.42

Innovator Versus Generic Antiepileptic Drug Evaluation

A summary of search results is presented in Figure 3.

This figure shows the flow of study identification and selection. The initial database search resulted in 321 records from MEDLINE, 34 records from CENTRAL and 25 records from WEB OF SCIENCE. An additional nine records were identified from other sources. After duplicates were removed, there were 356 unique citations eligible for title and abstract screening. The first phase of screening excluded 267 records for the following reasons: 150 records were not reports of new discovery in humans, 59 records were not in epilepsy patient population, and another 58 records did not compare innovator to generic AED. This process left 89 records to assess for eligibility by screening the full-text articles. The second phase of screening excluded 18 articles because 13 articles were not reports of new discovery in humans and 2 articles were not in epilepsy patient population, 2 articles did not evaluate outcomes of interest, and one article was not published in the English language. Seventy one articles were included in qualitative synthesis. Of these, 18 trials were eligible for quantitative synthesis.

Figure 3

PRISMA flow diagram of citations identified and evaluated in the innovator versus generic search. AED = antiepileptic drug; PRISMA = preferred reporting items for systematic reviews and meta-analyses

Upon conducting the literature search to identify articles that compared innovator antiepileptic medications to its generic version, we retrieved 380 unique citations, and another 9 citations were identified from other sources. After removing duplicates, 356 citations were retrieved. Two hundred sixty-seven articles were excluded during the title and abstract review, and 18 articles were excluded during the full-text review. A list of articles excluded during full text review can be found in Appendix C. A total of 71 articles were found to match our inclusion criteria. Although no language restrictions were imposed in the literature search, one article was excluded from the full text review because it was not published in the English language. Given the small number of articles excluded, their small sample sizes, and the traditional medicine versus complementary medicine topic, we do not anticipate that their inclusion would have impacted our results in a meaningful way.42

General Overview of Results

Table 10 delineates the results of the CER with a rating of the strength of evidence. The sections that follow provide much more detail into the analyses conducted and the nuances necessary to understand the analyses and how they apply to health care decisionmakers, clinicians, and patients.

Key Question 1. In patients with epilepsy, what is the comparative effectiveness/efficacy of antiepileptic medications on health outcomes: mortality, hospitalizations, office/emergency department visits, composite endpoint (ambulance services, hospitalizations, or emergency department visits for epilepsy), health-related quality of life, seizures (time to first seizure, time to exit from trial due to lack of efficacy, proportion of seizure-free patients, proportion of patients with seizure remission, breakthrough seizures, frequency of seizures), secondary seizure injury (fracture, laceration, head trauma, aspiration pneumonia), status epilepticus, loss of driver's license, and loss of employment?

Key Points

  • There was no difference observed in the risk of mortality when newer antiepileptic medications were compared with carbamazepine, phenytoin, or valproic acid. No other comparisons between newer and older antiepileptic medications were available.
  • No difference in the risk of maintaining seizure freedom was seen when newer antiepileptic medications were compared versus carbamazepine, controlled/sustained release carbamazepine, phenytoin, or valproic acid in controlled clinical trials.
  • The risk of seizure freedom at 12 or 24 months was significantly lower for newer antiepileptic agents versus carbamazepine; therefore, patients were more likely to be seizure free at 12 or 24 months when receiving carbamazepine compared with newer agents. When either gabapentin or oxcarbazepine was compared to carbamazepine individually, the risk of seizure freedom was significantly reduced at 12 and 24 months; therefore, patients were more likely to be seizure free at 12 and 24 months when receiving carbamazepine compared with gabapentin or oxcarbazepine. When topiramate was compared with carbamazepine, the risk of seizure freedom at 12 months was significantly reduced; therefore, patients were more likely to be seizure free at 12 months when receiving carbamazepine. No differences in 12- or 24-month seizure freedom were seen for newer antiepileptic medications versus valproic acid although this was based on a single controlled clinical trial. No trials were available evaluating newer antiepileptic medications versus controlled/sustained-release carbamazepine or phenytoin.
  • There was a significant increase in the time to first seizure when newer antiepileptic medications were compared versus phenytoin. No difference in the time to first seizure was seen between newer antiepileptic medications versus carbamazepine or valproic acid. However in individual newer agent versus carbamazepine or valproic acid analyses, significant reductions were seen for gabapentin and vigabatrin versus carbamazepine and for lamotrigine versus valproic acid.
  • Five instruments were used to assess for health-related quality of life in the newer versus older antiepileptic medication evaluation. The instruments have differences in the importance of subscales as well as the areas that are evaluated.
    • In a direct comparative trial, gabapentin, lamotrigine, oxcarbazepine, and topiramate similarly impacted health-related quality of life versus carbamazepine. Tiagabine and carbamazepine both significantly improved subscales of health-related quality of life versus baseline in one trial, but lamotrigine significantly improved health-related quality of life versus baseline while carbamazepine did not in another trial.
    • In one direct comparative trial, there was no significant difference in the median change from baseline in health-related quality of life when lamtrigine was compared to sustained-release carbamazepine.
    • In a direct comparative trial, lamotrigine had a greater positive impact on health-related quality of life versus phenytoin. Tiagabine and phenytoin both significantly improved a subscale of health-related quality of life versus baseline in one trial.
    • In one direct comparative trial, lamotrigine and topiramate yielded a similar effect to valproic acid and in another trial topiramate and valproic acid had similar effects on health-related quality of life. In another trial, lamotrigine had a greater positive effect on three subscales of health-related quality of life versus valproic acid.
  • In controlled clinical trials, the available data on seizure occurrence or frequency for innovator and generic antiepileptic medications is confined to carbamazepine and to a lesser extent phenytoin and valproic acid.
  • For a population of people, seizure occurrence, seizure frequency, withdrawals for any reason, and withdrawals due to lack of efficacy are similar during periods in which innovator and generic versions of antiepileptic agents are used.
    • The impact of switching patients stabilized on an innovator or generic version to an alternative version on these endpoints cannot be answered by the available controlled clinical trials or observational studies.
  • Important final health outcomes including mortality, health-related quality of life, loss of driver's license or employment, time to first seizure, seizure remission, secondary seizure injury, and status epilepticus could not be assessed.
  • Controlled observational studies suggest that switching from one version of an antiepileptic medication (either innovator or generic) to another version may increase medical service utilization.

Detailed Analysis

Older Versus Newer Study Design and Population Characteristics

Forty-eight controlled trial reports (n=13,039) evaluated older versus newer antiepileptic medication comparisons and were eligible for inclusion (Appendix F Table 5).47-68,68-93

Twenty-four reports were funded by the pharmaceutical industry,49,51,52,59,60,63,64,66-69,71-74,76-78,80-83,86,90,93 5 reports were funded by government or foundation funding,78,85,89,91,92 and funding was not reported for 19 reports.47,48,50,53,55-58,61,62,65,66,70,75,84,87,94

Thirteen reports were conducted in the United States;50,51,66,68,69,71,72,76,78,84,89,90,93 two reports were conducted in China;92,94 two reports were conducted in Finland;47,53 one report was conducted in Germany;80 three reports were conducted in the United Kingdom;61,85,87 three reports were conducted in Italy;56,75,91 one report was conducted in Korea;83 one report was conducted in the Netherlands;65 eight multinational reports were conducted in Europe;49,52,63,64,67,70,82,86 one multinational report was conducted in Asia, Europe, North America, and South America;88 one multinational report was conducted in Australia and Europe;55 one multinational report was conducted in Australia, Europe, and South Africa;62 one multinational report was conducted in Australia, Europe, South Africa, the United States, and South America;74 and one multinational report was conducted in Europe and South Africa.81 Three reports did not report any country.48,73,77

Baseline characteristics are presented in Appendix F Tables 59. The average age is between 12 years and 77 years, and the percentage of male participants ranged from no male participants to 100 percent. Body weight ranged between 19 and 83.5 kg within the 15 trials that reported it.49,51,56-59,61,65,70,72,78,81,83,86,88,90 Ethnicity was reported by 10 trials.51,57,59,68,72,78,81,88-90 The percentage of Caucasian participants ranged from 47 to 92.1 in the 10 trials that reported the percentage of Caucasian participants,51,57,59,68,72,78,81,88-90 the percentage of Black patients ranged from 3 to 27.6 in the 10 trials that reported the percentage of Black participants,51,57,59,68,72,78,81,88-90 the percentage of Asian participants ranged from 0 to 11 in the 4 trials that reported the percentage of Asian participants,68,72,81,90 the percentage of Hispanic participants ranged from 0 to 8 in the four trials that reported the percentage of Hispanic participants,68,72,78,88 and the percentage of participants with another ethnicity ranged from 1.5 to 34 in the 7 trials that reported participants with another ethnicity.51,57,59,78,81,89,90

Epilepsy history was only reported by 23 trials,50,51,55,57-60,62,65,68-70,72,75,76,80,82,83,85,86,88-90 and only 30 trials reported seizure type.47,49,51-53,56-60,63,64,68-70,72,74-78,80-82,84,85,88-90,94

Fourteen trials reported the number of patients that were untreated with antiepileptic agents prior to enrollment.52,53,56,57,63,64,70,72,74,78,82,84,85,88 Prior or concurrent use of carbamazepine, phenytoin, valproic acid, phenobarbital, gabapentin, lorazepam, or other therapy was reported in 10 trials.47,62,68,69,72,76,84,85,88,95.

Fifteen observational reports (n=2469) evaluated older versus newer antiepileptic medication comparisons and were eligible for inclusion (Appendix F Table 5).54,54,79,95-106 Three reports were funded by pharmaceutical companies,96,97,103 five reports were funded by government or foundations, and54,102,105,106 seven reports did not report the role of funding.79,95,98-101,104 Three reports were conducted in the United States,96,97,102,103 one report was conducted in Denmark,54 one report was conducted in Italy,98 one report was conducted in Scotland,99 one report was conducted in Hungary,95 one report was conducted in Poland,79 one report was conducted in Turkey,100 two reports were conducted in Korea,101,104 one report was conducted in China,105 and one report was conducted in Brazil.106 (Appendix F Table 5)

Baseline characteristics are elucidated in Appendix F Tables 59. The average age ranged between 8.5 months and 36.9 years, and the percentage of male participants ranged from 25 to 60 percent. Body weight ranged from 38.15 to 73 kg in the two studies that reported body weight,100,102 and the percentage of Caucasian participants ranged from 61 to 76 percent in the study that reported the percentage of Caucasian participants (Appendix F Table 7).102 Epilepsy history was reported in two observational studies,99,102 and only three of the studies reported the seizure type (Appendix F Table 8).79,98,100

Four studies reported the number of patients that were untreated with antiepileptic agents prior to enrollment.54,98,99,101 Prior or concurrent use of carbamazepine, phenytoin, valproic acid, phenobarbital, gabapentin, lorazepam, or other therapy use was reported in four studies (Appendix F Table 9).79,95,100,102

Descriptive studies were not included in the older versus newer antiepileptic medication evaluation.

Innovator Versus Generic Study Design and Population Characteristics

Seventeen controlled trial reports (n=365) evaluated innovator to generic or generic to innovator comparisons and were eligible for inclusion (Appendix F Table 1).107-123 Two phenytoin trials that appeared otherwise eligible for our analyses were excluded upon detailed evaluation.124,125 In the first trial, the products were coded by letter, and it could not be determined which products, if any, were innovator or generic. In the second trial, two innovator products from the same manufacturer but manufactured in different countries were compared with each other, and the generic versions were not of the same salt form (Appendix C Table 1).124,125

Only 1 of the 17 reports specified that they were limited to “A” rated versions of the generic medications.112 Four were funded by the pharmaceutical industry,107,110,112,119 and funding was not known for eight studies.111,113,115,116,118,120-122 Multiple studies were included in one of the reports.112 Four reports were conducted in the United States,109,112,121,123 five were conducted in the United Kingdom,110,111,118,119,122 two in Finland107,113 and one each in Germany, Thailand, Netherlands, Denmark, Sweden and India (Appendix F Table 1).108,114-117,120

Baseline characteristics are presented in Appendix F Tables 14. The average age ranged between 9.5 and 45.1 years, and the percentage of male participants ranged from 35 to 100 percent. Body weight ranged between 35.8 and 59.6 kg within the six studies that reported it.108,110,114,120,122,123 Ethnicity was not reported in any study (Appendix F Table 2). Epilepsy history was only reported by three trials (either new-onset or chronic epilepsy),110,115,120 and only six of the studies reported the seizure type (Appendix F Table 3).110,113-116,119

Patients in these trials had been previously treated with antiepileptic medications. Prior or concurrent use of carbamazepine, phenytoin, or valproic acid was reported in 11 out of 16 trials (Appendix F Table 4).107-110,112-116,118,121 The use of combination therapy was reported in 8107-109,112,114-116,121 of the 17 trials and ranged between 0 and 80 percent.

Nine controlled observational reports (n=61,684), not constituted by patients in clinical trials, evaluated innovator to generic, generic to generic, or generic to innovator switches and were eligible for inclusion (Appendix F Table 1).33,126-133 Only three of the eight reports specified that they were limited to “A” rated versions of the generic medications.126-128 Eight of the reports were funded by the pharmaceutical industry,33,126,127,129-133 and one was funded by a health insurance provider.128 Multiple studies were included in three of the reports.33,129,131 Five reports were conducted in the United States126-130 and the other four in Canada (Appendix F Table 1).33,131-133

Baseline characteristics are elucidated in Appendix F Tables 23. The average age ranged between 33.7 and 52.5 years and the percentage of male participants ranged from 32.3 to 50.8 percent. Body weight and ethnicity were not reported in these observational studies (Appendix F Table 2). Epilepsy history was not reported in the observational studies, and only three of the studies reported the seizure type (either partial or generalized) (Appendix F Table 3).126-128

Patients in observational studies had been previously treated with antiepileptic medications (Appendix F Table 4).33,126-133 Prior or concurrent use of carbamazepine, phenytoin, or valproic acid was not reported. The use of combination therapy was reported in five33,128,131-133 of the eight trials, and the percent of patients on combination treatment ranged between 52 and 94 percent.

Twenty-nine descriptive reports (n=2,190) assessed innovator to generic, generic to generic, or generic to innovator switches. Seventeen (60.7 percent) of these descriptive studies were case reports or case series, four (14.3 percent) were pure surveys, one (3.6 percent) was a survey and a case series, and seven were other designs where patients were described but not compared with a control group.

Four of the reports were funded by the pharmaceutical industry,134-137 two of the reports were funded by the National Institutes of Health,138,139 and funding for all the other reports was not specified.116,140-161 Multiple descriptive studies were included in 10 of the reports.134-136,138,140-145 Fifteen reports were conducted in the United States,135,136,138,139,141,143-149,155,158,159 two reports were conducted in the United Kingdom,134,140 two reports were conducted in Poland,142,156 two reports were conducted in Germany153,157 and one report each was conducted in Switzerland, Italy, Canada, and Denmark.116,137,150,152 Study location for four other reports was unknown (Appendix F Table 1).151,154,160,161

Baseline characteristics are reported in Appendix F Tables 2-3. The average age ranged between 6.0 and 66.0 years, and the percentage of male participants ranged from 0 to 100 percent. Body weight and ethnicity were not reported in these reports (Appendix F Table 2). Epilepsy history was only reported by five studies.145,147-149,158 Epilepsy duration ranged from 1 to 16 years.145,149,158 Four studies reported chronic onset of epilepsy which was either at birth or 1 year of age.147-149,158 Twelve studies reported seizure type as partial, simple partial, complex partial, generalized, tonic-clonic or absence and the proportion of patients with each type of seizure varied between studies (Appendix F Table 3).116,136,140,143,146-148,150,156,158,160,161

Patients in fourteen studies had been previously treated with antiepileptic medications (Table 4).116,143,145-152,156-158,160 Prior or concurrent use of carbamazepine, phenytoin, or valproic acid was reported in eight studies.116,146-150,152,160 The use of combination therapy was reported in 12 of these trials116,143,145,147-152,156-158 and ranged between 21 and 100 percent.

Given the inherent limitations and biases associated with uncontrolled data, and the specific limitations associated with these reports in particular, these studies will not be discussed in the results section of this report.

Outcome Evaluations

Mortality
Older Versus Newer Antiepileptic Drug Evaluation
Carbamazepine Versus Newer

Six randomized controlled trials reported the number of patients who died when receiving a newer antiepileptic compared with carbamazepine, and all six were amenable for pooling.55,62,63,76,78,85

Two randomized controlled trials reported the number of patients who died when gabapentin was compared with carbamazepine, and both were amenable for pooling.78,85 The risk of death was nonsignificantly decreased by 8 percent when gabapentin was compared with carbamazepine (relative risk [RR] 0.92 [0.57 to 1.48]) (Appendix J Figure 1).

Five randomized controlled trials reported the number of patients who died when lamotrigine was compared with carbamazepine, and all five were amenable for pooling.55,62,76,78,85 The risk of death was nonsignificantly decreased by 37 percent when lamotrigine was compared with carbamazepine (RR 0.63 [0.37 to 1.04]) (Appendix J Figure 1). No statistical heterogeneity (I2: 0 percent) or publication bias was detected (Egger's: p=0.682).

One randomized controlled trial reported the number of patients who died when oxcarbazepine was compared with carbamazepine.85 The risk of death was nonsignificantly decreased by 50 percent when oxcarbazepine was compared with carbamazepine (RR 0.50 [0.19 to 1.27]) (Appendix J Figure 1).

One randomized controlled trial reported the number of patients who died when topiramate was compared with carbamazepine.85 The risk of death was nonsignificantly decreased by 6 percent when topiramate was compared with carbamazepine (RR 0.94 [0.50 to 1.78]) (Appendix J Figure 1).

One randomized controlled trial reported the number of patients who died when vigabatrin was compared with carbamazepine.63 The risk of death was nonsignificantly increased by 2.0-fold when vigabatrin was compared with carbamazepine (RR 2.01 [0.26 to 15.27]) (Appendix J Figure 1).

Six randomized controlled trials reported the number of patients who died when either gabapentin, lamotrigine, oxcarbazepine, or topiramate was compared with carbamazepine and all six were amenable for pooling.55,62,63,76,78,85 The risk of death was nonsignificantly decreased by 25 percent when all newer antiepileptics were compared with carbamazepine (RR 0.75 [0.51 to 1.12]) (Appendix J Figure 1). No statistical heterogeneity (I2: 0 percent) or publication bias was detected (Egger's: p=0.598).

Phenytoin Versus Newer

Three randomized controlled trials reported the number of patients who died when newer antiepileptics were compared with phenytoin and all three were amenable for pooling.57,90

One randomized controlled trial reported the number of patients who died when lamotrigine was compared with phenytoin.64 The risk of death was nonsignificantly decreased by 78 percent when lamotrigine was compared with phenytoin (RR 0.22 [0.02 to 2.41]) (Appendix J Figure 2).

One randomized controlled trial reported the number of patients who died when oxcarbazepine was compared with phenytoin.57 The risk of death was nonsignificantly decreased by 80 percent when oxcarbazepine was compared with phenytoin (RR 0.20 [0.02 to 2.22]) (Appendix J Figure 2).

One randomized controlled trial reported the number of patients who died when topiramate was compared with phenytoin.90 The risk of death was nonsignificantly decreased by 4 percent when topiramate was compared with phenytoin (RR 0.96 [0.06 to 16.60]) (Appendix J Figure 2).

Three randomized controlled trials reported the number of patients who died when either lamotrigine, oxcarbazepine or topiramate was compared with phenytoin and all three were amenable for pooling.57,64 The risk of death was nonsignificantly decreased by 70 percent when the newer agents were compared with phenytoin (RR 0.30 [0.05 to 1.95]) (Appendix J Figure 2). No significant statistical heterogeneity was detected (I2: 0 percent); however, tests for publication bias could not be performed.

Valproic Acid Versus Newer

Three randomized controlled trials reported the number of patients who died when newer antiepileptics were compared with valproic acid and all three were amenable for pooling.61,76,85

Two randomized controlled trials reported the number of patients who died when lamotrigine was compared with valproic acid and both were amenable for pooling.76,85 The risk of death was nonsignificantly increased by 6 percent when lamotrigine was compared with valproic acid (RR 1.06 [0.30 to 3.76]) (Appendix J Figure 3).

One randomized controlled trial reported the number of patients who died when topiramate was compared with valproic acid.85 The risk of death was nonsignificantly decreased by 25 percent when topiramate was compared with valproic acid (RR 0.75 [0.19 to 2.95]) (Appendix J Figure 3).

One randomized controlled trial reported the number of patients who died when vigabatrin was compared with valproic acid.62 The risk of death was nonsignificantly decreased by 1 percent when vigabatrin was compared with valproic acid (RR 0.99 [0.06 to 17.08]) (Appendix J Figure 3).

Three randomized controlled trials reported the number of patients who died when the newer antiepileptics lamotrigine, topiramate or vigabatrin were compared with valproic acid and all three were amenable for pooling.62,76,85 The risk of death was nonsignificantly decreased by 6 percent when newer agents were compared with valproic acid (RR 0.94 [0.31 to 2.80]) (Appendix J Figure 3). No publication bias (Egger's p=0.448) was detected.

Phenobarbital or Primidone Versus Newer

No data are available for this comparison.

Ethosuximide Versus Newer

No data are available for this comparison.

Innovator Versus Generic Antiepileptic Drug Evaluation

None of the available controlled clinical trials or observational studies reported on mortality as an endpoint.

Use of Medical Services
Older Versus Newer Antiepileptic Drug Evaluation

No data are available comparing newer versus older antiepileptic medications for any endpoint in this section.

Innovator Versus Generic Antiepileptic Drug Evaluation
Office or Emergency Room Visits

No controlled clinical trials and four controlled observational reports (n=3,852) evaluated innovator to generic, generic to generic, or generic to innovator switches and were eligible for inclusion and reported on at least one of the following endpoints: outpatient visits, hospitalizations, or hospital stay duration (Appendix G Table 13).33,129,132,133 None of the four reports specified that they were limited to “A” rated versions of the generic medications. The pharmaceutical industry funded all of the reports.33,129,132,133 Multiple studies were included in two of the reports.33,129 One report was conducted in the United States129 and the other three in Canada (Appendix G Table 13).33,132,133

Baseline characteristics are elucidated in Appendix F Table 2, Appendix F Table 3. The average age ranged between 33.7 and 52.5 years, and the percentage of male participants ranged from 32.3 percent to 50.8 percent. Body weight and ethnicity were not reported in these observational studies (Appendix F Table 2). Epilepsy history or seizure type was also not reported in these observational studies (Appendix F Table 3).

Patients in these studies had been previously treated with antiepileptic medications (Appendix F Table 4).33,129,132,133 Prior or concurrent use of carbamazepine, phenytoin, or valproic acid was not well reported. The use of combination therapy was reported in three33,132,133 of the four trials and ranged between 70 percent and 94 percent.

No controlled clinical trials and three controlled observational reports (n=17,424) evaluated innovator to generic, generic to generic, or generic to innovator switches and were eligible for inclusion and reported on a composite endpoint of medical service utilization.126-128 All of the observational studies were conducted in the United States, two were funded by Abbott Laboratories and employed a similar methodology, and one was funded by Express Scripts. Baseline characteristics are elucidated in Appendix F Table 2, Appendix F Table 3. The average age ranged between 35.6 and 44 years, and the percentage of male participants ranged from 43.9 to 49 percent. Body weight and ethnicity were not reported in these observational studies (Appendix F Table 2). Epilepsy history or seizure type was also not reported in these observational studies (Appendix F Table 3).

Ambulance Services

There were no controlled clinical trials or observational studies that met our inclusion criteria that evaluated ambulance services as a sole endpoint.

Outpatient Medical Care Utilization

No controlled clinical trials and four large observational studies evaluated the impact of switching from innovator to generic, generic to innovator, or generic to generic antiepileptic medication33,129,133 on office or emergency room visits. One was conducted in the United States, and the other three studies were conducted in Canada. Two of the studies were supported by GlaxoSmithKline33,129 and the other two studies were supported by Ortho-McNeil-Janssen.132,133 All four studies employed similar methodology, but none of the studies specified whether they were limited to “A” rated generic products. These trials are not amenable to statistical pooling and are discussed qualitatively.

In the first study, a retrospective open-cohort design was used to classify the duration of observations into two mutually exclusive periods of innovator and generic use of antiepileptic drugs.129 Patients with epilepsy who had an ambulance ride, emergency department visit, or office visit between January 1, 2000, and October 31, 2007, were compared during innovator and generic use periods using a person-time approach. Incidence rates of outpatient visits (office and emergency room visits) were calculated for antiepileptic drugs. The duration of prescriptions was normalized to 28 days to enable incidence rate comparisons. Study results were further stratified into stable versus unstable epilepsy groups. The stable group was defined as patients with ≤ 2 outpatient services per year on average throughout the observation period and no emergency room visits associated with epilepsy or nonfebrile convulsions. All other patients were defined as having unstable epilepsy. For epilepsy-related medical resource use, observed differences in outpatient visits in both the stable and unstable group showed higher utilization rates during generic-use periods (2.52 [innovator] versus 2.92 [generic]; adjusted incidence rate ratio [IRR] 1.20 [1.19 to 1.21]) in the stable patient group and (23.33 [innovator] versus 28.36 [generic]; adjusted IRR 1.16 [1.16 to 1.17]) in the unstable patient group.

In the second study, a retrospective open-cohort design was used in patients with epilepsy who had any outpatient visits between April 1, 1998, and July 31, 2006. Incidence rates of outpatient visits were calculated and compared between periods of innovator versus generic use of lamotrigine.33 Incidence rates were calculated as the number of events divided by the number of person-years of observation. To account for varying days of supply associated with different dispensations, the dispensation length was set to 28 days. Outpatient visits were more frequent during the generic period compared with the innovator periods (9.25 versus 8.24 visits per person per year; RR 1.13 [1.09 to 1.18]; p < 0.0001).

In the third study, a retrospective open-cohort design was used to classify the observation period into mutually exclusive periods of innovator topiramate use only, single-generic topiramate use, and multiple-generic topiramate use.132 Incidence rates of outpatient visits were calculated and compared between periods of innovator versus single-generic and multiple-generic use of topiramate. Incidence rates were calculated as the number of events divided by the number of person-years observed. The days of drug supplied were harmonized to 28 days to enable incidence rate comparisons. Incidence rates for outcomes during the innovator period versus single-generic and multiple-generic periods were compared using incidence rate ratios. Outpatient visits showed no significant differences among the three studied periods (9.07 [innovator] versus 9.48 [single-generic] versus 8.74 [multiple-generic] visits per person per year; adjusted IRR 0.99 (0.94 to 1.04) for single-generic versus innovator and adjusted IRR 0.95 (0.88 to 1.02) for multiple-generic versus innovator).

In the fourth study, a retrospective open-cohort design was used to classify the observation period into mutually exclusive periods of innovator and generic use of topiramate.133 Incidence rates of outpatient visits were expressed as frequency per person per year, and were calculated as the number of events divided by the number of person-years observed, and were compared between periods of innovator versus generic use of topiramate using IRRs. The days of drug supplied were harmonized to 28 days to enable incidence rate comparisons. Outpatient visits showed no significant difference among the two periods (9.0 [innovator] versus 9.1 [generic] visits per person per year; adjusted IRR 0.99 [0.96 to 1.03] for innovator versus generic).

Hospitalizations

No controlled trials and four large observational studies evaluated the impact of switching from innovator to generic, generic to innovator, or generic to generic antiepileptic medication33,129,132,133 on hospitalizations. One study129 was conducted in United States and the other three studies were conducted in Canada. Two of the studies were supported by GlaxoSmithKline33,129 and the other two studies were supported by Ortho-McNeil-Janssen.132,133 All four studies employed similar methodology, but none of the studies specified whether they were limited to “A” rated generic products. These trials are not amenable for statistical pooling and are discussed qualitatively.

In the first study, a retrospective open-cohort design was used to classify the duration of observations into two mutually exclusive periods of innovator and generic use of antiepileptic drugs. Patients with epilepsy who were hospitalized between January 1, 2000, and October 31, 2007, were compared during innovator and generic use periods using a person-time approach. Incidence rates of hospitalizations were calculated for antiepileptic drugs.129 The duration of prescriptions was normalized to 28 days to enable incidence rate comparisons. Study results were further stratified into stable versus unstable epilepsy groups. The stable group was defined as patients with ≤ 2 outpatient services per year on average throughout the observation period and no ER visit associated with epilepsy or nonfebrile convulsions. All other patients were defined as having unstable epilepsy. For epilepsy-related medical resource use, observed differences in hospitalization in both the stable and unstable group showed higher utilization rates during generic-use periods (0.05 [innovator] versus 0.06 [generic]; adjusted IRR [incidence rate ratio] 1.31 [1.24 to 1.40]) in stable patient group and (0.34 [innovator] versus 0.47 [generic]; adjusted IRR 1.30 [1.25 to 1.36]) in unstable patient group.

In the second study, a retrospective open-cohort design was used in patients with epilepsy who had any outpatient visits between April 1, 1998, and July 31, 2006. Incidence rates of hospitalizations were calculated and compared between periods of innovator versus generic use of lamotrigine.33 Incidence rates were calculated as the number of events divided by the number of person-years of observation. To account for varying days of supply associated with different dispensations, the dispensation length was set to 28 days. Rates of inpatient hospitalizations were not statistically different between the generic and innovator periods (0.56 versus 0.49 visits per person per year; RR 1.14 [0.96 to 1.35]; p = 0.1264).

In the third study, a retrospective open-cohort design was used to classify the observation period into mutually exclusive periods of innovator topiramate use, single-generic topiramate use, and multiple-generic topiramate use.132 Incidence rates of hospitalizations were calculated and compared between periods of innovator versus single-generic and multiple-generic use of topiramate. Incidence rates were calculated as the number of events divided by the number of person-years observed. The days of drug supplied were harmonized to 28 days to enable incidence rate comparisons. Incidence rates for outcomes during the innovator period versus single-generic and multiple-generic periods were compared using IRRs. After covariate adjustment, multiple generic use was associated with significantly higher incidence of hospitalization relative to innovator-only use, while the difference between single-generic and innovator periods was not significant (0.48 [innovator] versus 0.52 [single-generic] versus 0.83 [multiple-generic] hospitalizations per person per year; adjusted IRR 1.08 [0.88 to 1.34] for single-generic versus innovator and adjusted IRR 1.65 [1.28 to 2.13] for multiple-generic versus innovator).

In the fourth study, a retrospective open-cohort design was used to classify the observation period into mutually exclusive periods of innovator and generic use of topiramate.133 Incidence rates of hospitalizations were expressed as frequency per person per year, and were calculated as the number of events divided by the number of person-years observed, and were compared between periods of innovatored versus generic use of topiramate using IRRs. The days of drug supplied were harmonized to 28 days to enable incidence rate comparisons. Rates of hospitalization was significantly higher in generic use period compared with innovator periods (0.5 [innovator] versus 0.6 [generic] visits per person per year; adjusted IRR 1.17 [1.03 to 1.33]).

Hospital Stay Duration
Innovator Versus Generic Antiepileptic Drug Evaluation

No controlled trials and four large observational studies evaluated the impact of switching from innovator to generic, generic to innovator, or generic to generic antiepileptic medication33,129,133,162 on office or emergency room visits. One study129 was conducted in United States and the other three studies were conducted in Canada. Two of the studies were supported by GlaxoSmithKline33,129 and the other two studies were supported by Ortho-McNeil-Janssen.132,133 All four studies employed similar methodology, but none of the studies specified whether they were limited to “A” rated generic products. These studies were not amenable to statistical pooling and are discussed qualitatively.

In the first study, a retrospective open-cohort design was used to classify the duration of observations into two mutually exclusive periods of innovator and generic use of antiepileptic drugs. Hospital duration of patients with epilepsy who were hospitalized between January 1, 2000, and October 31, 2007, was compared during innovator and generic use periods. Incidence rates of length of hospital stay was calculated for antiepileptic drugs.129 The duration of prescriptions was normalized to 28 days to enable incidence rate comparisons. Study results were further stratified into stable versus unstable epilepsy groups. The stable group was defined as patients with ≤ 2 outpatient services per year on average throughout the observation period and no emergency room visit associated with epilepsy or nonfebrile convulsions. All other patients were defined as having unstable epilepsy. For epilepsy-related medical resource use, hospital stays lasted significantly longer on average during generic antiepileptic drug treatment in both stable and unstable group (1.02 [innovator] versus 1.38 [generic]; adjusted IRR 1.33 [1.30 to 1.36]) in stable patient group and (2.33 [innovator] versus 3.29 [generic]; adjusted IRR 1.34 [1.32 to 1.36]) in unstable patient group.

In the second study, a retrospective open-cohort design was used in patients with epilepsy who had any hospitalizations between April 1, 1998, and July 31, 2006. Incidence rates of length of hospital stay was calculated and compared between periods of innovator versus generic use of lamotrigine,33 Incidence rates were calculated as the number of days divided by the number of person-years of observation. To account for varying days of supply associated with different dispensations, the dispensation length was set to 28 days. The average length of hospital stay was longer during the generic period compared with the innovator periods (4.86 versus 3.29 days per patient per year; RR 1.48 [confidence interval not reported]; p < 0.0001).

In the third study, a retrospective open-cohort design was used to classify the observation period into mutually exclusive periods of innovator topiramate only, single-generic topiramate use, and multiple-generic topiramate use.132 Incidence rates of length of hospital stay was calculated and compared between periods of innovator versus single-generic and multiple-generic use of topiramate. Incidence rates were calculated as the number of days divided by the number of person-years observed. The days of drug supplied were harmonized to 28 days to enable incidence rate comparisons. Incidence rates for outcomes during the innovator period versus single-generic and multiple-generic periods were compared using IRRs. Significantly longer mean hospital lengths of stay were observed for multiple-generic period and for single-generic period than for innovator period (2.55 [innovator] versus 3.22 [single-generic] versus 3.88 [multiple-generic] days per person per year; adjusted IRR 1.12 [1.03 to 1.23] for single-generic versus innovator and adjusted IRR 1.43 [1.27 to 1.60] for multiple-generic versus innovator).

In the fourth study, a retrospective open-cohort design was used to classify the observation period into mutually exclusive periods of innovator and generic use of topiramate.133 Incidence rates of length of hospital stay was expressed as days per person per year, and were calculated as the number of days divided by the number of person-years observed, and were compared between periods of innovatored versus generic use of topiramate using IRRs. The days of drug supplied were harmonized to 28 days to enable incidence rate comparisons. Significantly longer average length of hospital stays were observed during generic period compared with innovator period (2.4 [innovator] versus 3.1 [generic] days per person per year; adjusted IRR 1.21 [1.15 to 1.28]).

Composite of Medical Services
Innovator Versus Generic Antiepileptic Drug Evaluation

No controlled trials and three large observational studies evaluated the impact of switching from innovator to generic, generic to innovator, or generic to generic antiepileptic medication on a composite of medical service utilization.126-128 All three of the observational studies were conducted in the United States, employed similar methodology, and were limited to “A” rated products. The first two were supported by Abbott Laboratories126,127 while the last was supported by the pharmacy benefit managing company Express Scripts, Inc.128

In the first study, patients with epilepsy who had an ambulance ride, emergency department visit, or inpatient hospitalization between July 1, 2006, and December 31, 2006, without a previous acute event in the past 6 months were defined as cases, while those receiving an office visit for epilepsy during the same time period were defined as controls.126 The percentage of patients switching from an “A” rated antiepileptic medication to another “A” rated version of the same antiepileptic medication was compared between groups. This could have been switching from innovator to generic, generic to innovator, or generic to generic. Patients were matched 1:3 (cases:controls) for diagnosis and age. Diagnosis was divided into seizure type (generalized, partial, or other) and whether the seizure type was intractable or not. Patients in the case group and control group were well matched for seizure type and severity, male gender, age, and region of the country. The cases were more likely to be insured by Medicaid than the controls. The cases were significantly more likely to have undergone a switch from one “A” rated antiepileptic medication to another “A” rated version of the medication in the base case analysis (odds ratio [OR] 1.81 [1.25, 2.63], 11.3 percent versus 6.5 percent), the analysis excluding patients with a concurrent change in dosage (OR 2.01 [1.19, 3.40], 9.7 percent versus 5.1 percent), and the analysis excluding patients with Medicaid coverage (OR 1.86 [1.26, 2.73], 11.3 percent versus 6.4 percent). However, these results are unadjusted and therefore may be biased.

In the second study, patients with epilepsy who received an ambulance service, emergency department visit, or inpatient hospitalization between October 1, 2005, and December 31, 2006, without a previous acute event in the past 6 months were cases while those receiving an office visit for epilepsy during the same time period were controls.127 The percentage of patients switching from an “A” rated antiepileptic medication to another “A” rated version of the same antiepileptic medication was compared between groups. This could have been from innovator to generic, generic to innovator, or generic to generic. Patients were matched 1:3 (cases:controls) for gender, age, and diagnosis. Diagnosis was divided into seizure type (generalized, partial, or other) and whether the seizure type was intractable or not. Patients in the case and control group were well matched for age, male gender, and seizure type and severity. The cases were more likely to be insured by Medicaid than the controls, and the regional distribution between cases and controls was different. The cases were significantly more likely to have undergone a switch from one “A” rated antiepileptic medication to another “A” rated version of the medication in the base case analysis (OR 1.84 [1.44, 2.36], 11.0 percent versus 6.3 percent]), analysis excluding patients with a concurrent change in dosage (OR 2.86 [2.13, 3.83], 11.5 percent versus 4.4 percent), and analysis excluding patients with Medicaid coverage (OR 1.83 [95 percent CI 1.41 to 2.37], 10.6 percent versus 6.1 percent). No analysis was conducted based on geographic location.

In the third study, patients with epilepsy who had an emergency department visit, or inpatient hospitalization between January 1, 2006, and December 31, 2007, without a previous acute event in the past 6 months were defined as cases while controls were from the same population and matched on baseline epilepsy diagnosis and followup time since January 1, 2006.128 The exposure was a switch between “A” rated antiepileptic drugs in the 90 days prior to the matching date. This could have been switching from innovator to generic, generic to innovator, or generic to generic. Each case was matched to three controls with the same baseline diagnosis code for epilepsy (the most recent medical claim prior to December 31, 2005), and a total time at risk greater than or equal to that of the index date of the case. Patients in the case and control group were well matched for seizure type and severity, gender, age, and region of the country. The authors also controlled for the following confounders: person's risk of epilepsy exacerbation, change in disease severity, drug interactions, poor adherence, and change in patient diagnosis. The unadjusted odds ratio between switch and epilepsy exacerbation was 1.51 (1.29–1.76). After adjusting for potential confounders, the odds ratio was non significant 1.08 (0.91–1.29). The time evaluated for the number of antiepileptic drugs, the addition of a new antiepileptic drug, and the addition of a new interacting medication were extended to 180 days. This alternate analysis resulted in an increase in the total number of “A” rated switches by 523, for a total of 1,286. Upon reanalysis, the adjusted odds ratio of acute epilepsy exacerbations were non-significantly increased to 1.14 (0.99–1.31).

Health-Related Quality of Life
Older Versus Newer Antiepileptic Drug Evaluation

Five instruments, Quality of Life in Epilepsy-89 (QOLIE-89), Side Effect and Life Satisfaction Inventory (SEALS Inventory), Newly Diagnosed Epilepsy Quality of Life (NEWQOL), European Descriptive Health Related Quality of Life States (EQ-5D), and the World Health Organization-5 Well-Being Index (WHO-5) were employed to evaluate health-related quality of life (HRQoL) when newer antiepileptic drugs were compared with older antiepileptic drugs. The QOLIE-89 scale is a self-administered health-related quality of life inventory for adults with epilepsy. The inventory includes contains 17 multi-item measures of overall quality of life including: emotional well-being, role limitations due to emotional problems, social support, social isolation, energy/fatigue, worry about seizure, medication effects, health discouragement, work/driving/social function, attention/concentration, language, memory, physical function, pain, role limitations due to physical problems, and health perceptions. Scores are calculated for the individual subscales and for the total with higher scores representing an improvement. The SEALS Inventory is a 38-question self-completed inventory that measures patient satisfaction with antiepileptic drug therapy. The 38 questions are divided into the following five subgroups of worry, temper, cognition, dysphoria, and tiredness. Each item is scored on a 4-point scale with 0 = never, 1 = occasionally, 2 = sometimes, 3 = many times. The score for each subscale is the total number of points for each of the items in the subscale. The total SEALS Inventory score is the sum of the scores of the five subscales. Lower scores indicate fewer symptoms and higher health-related quality of life. The NEWQOL is a 93-item self-administered battery that measures the quality of life in patients with new-onset epilepsy who are 16 years of age and older. The NEWQOL is made up of 93 items; 81 of the items comprise 8 multi-item subscales that measure several health parameters, including anxiety, depression, social activities, symptoms, locus of control/mastery, neuropsychological problems, social stigma, worry, and work. The remaining items are single-item subscales that include general health, number of seizures, social limitations, social support, self concept, ambition limitations, health transition, and general limitations. The EQ-5D is a descriptive system of health-related quality of life states with five dimensions including mobility, self-care, usual activities, pain/discomfort and anxiety/depression designed to measure health status. Each dimension has three levels of response based on severity including no problems, some or moderate problems, and extreme problems. The WHO-5 questionnaire assesses psychological well being using five positively worded items including, mood (good spirits, relaxation), vitality (being active and waking up fresh and rested), and general interests (being interested in things).

Carbamazepine Versus Newer

Four randomized controlled trials reported information regarding HRQoL when newer antiepileptic drugs were compared with carbamazepine using the QOLIE-89, SEALS Inventory, and NEWQOL and EQ-5D instruments (Appendix G Table 14).28,67,69,163

One trial reported information regarding HRQoL using the QOLIE-89 instrument when patients were receiving carbamazepine or tiagabine in addition to baseline phenytoin therapy.69 This trial reported the total score and the subscales with statistically significant changes from baseline in the epilepsy domain for responders in each treatment group. For patients who received carbamazepine in addition to baseline phenytoin therapy, there were statistically significant improvements from baseline in the work-driving-social-relations (p=0.004) and seizure worry (p=0.016) subscales. For patients who received tiagabine in addition to baseline phenytoin therapy, there were statistically significant improvements from baseline in the attention-concentration (p=0.002), memory (p=0.042), language (p=0.004), and seizure worry (p=0.03) subscales.

Two trials reported information regarding HRQoL versus baseline using the SEALS Inventory when patients received lamotrigine or carbamazepine or sustained-release carbamazepine.67,163 The first trial reported the mean total SEALS Inventory scores recorded at baseline by treatment group. Patients in the lamotrigine group had a statistically significant improvement in SEALS scores from baseline (p<0.001). Patients in the carbamazepine had no significant change in SEALS scores from baseline (p=0.394).150 The second trial reported the median difference in SEALS score from baseline to 40 weeks of treatment for the sustained-release carbamazepine and lamotrigine groups. The trial reported no statistically significant difference in median SEALS score from baseline between the sustained-release carbamazepine and lamotrigine groups (P=0.54).163

One trial reported information regarding HRQoL using the NEWQOL and EQ-5D inventories when newer antiepileptics, gabapentin, lamotrigine, oxcarbazepine, and topiramate were compared with carbamazepine.28 The trial reported that there was no significant difference in anxiety, depression, adverse events profile, neurotoxicity, or global quality of life at 2 years when patients were receiving newer antiepileptics gabapentin, lamotrigine, oxcarbazepine, and topiramate versus patients receiving carbamazepine. The trial also reported that there was no significant difference in health related quality of life at 2 years when the newer antiepileptics gabapentin, lamotrigine, oxcarbazepine, and topiramate, were compared with carbamazepine using the EQ-5D.

One observational study reported information regarding HRQol using the WHO-5 question questionnaire for patients who were not pretreated with any antiepileptic drug and who received either carbamazepine or levetiracetam.164 The study reported that 31.7 percent of the patients treated with carbamazepine had an improvement in quality of life according to the WHO-5, while 4.9 percent had a decline in quality of life. In contrast, 21.9 percent of the patients treated with levetiracetam had an improvement in quality of life according to the WHO-5, while 1.4 percent had a decline in quality of life.

Phenytoin Versus Newer

Two randomized controlled trials reported information regarding HRQoL when newer antiepileptic drugs were compared with phenytoin using the QOLIE-89 scale the SEALS Inventory instruments (Appendix G Table 14).64,69

One trial reported information regarding HRQoL using the QOLIE-89 instrument when patients were receiving phenytoin or tiagabine in addition to baseline carbamazepine therapy.69 This trial reported the total score and the subscales with statistically significant changes from baseline in the epilepsy domain for responders in each treatment group. For patients who received phenytoin in addition to baseline carbamazepine therapy, there was a statistically significant improvement from baseline in the seizure worry subscale (p=0.007). For patients who received tiagabine in addition to baseline carbamazepine therapy, there were statistically significant improvements from baseline in the seizure worry (p=0.03) subscale.

One trial reported information regarding HRQoL using the SEALS Inventory when patients received lamotrigine compared with phenytoin.64 This trial reported the mean total SEALS Inventory scores by visit and the estimated difference between treatments in the overall change from baseline. The estimated difference between treatments in the overall change from baseline was 4.0 points greater for the lamotrigine group compared with the phenytoin group (p=0.02).

Valproic Acid Versus Newer

Three randomized controlled trials reported information regarding HRQoL when newer antiepileptic drugs were compared with valproic acid using the QOLIE-89, NEWQOL and EQ-5D instruments (Appendix G Table 14).28,29,71,73

One trial reported information regarding HRQoL using the NEWQOL, and EQ-5D inventories when the newer antiepileptic drugs lamotrigine and topiramate were compared with valproic acid.28 The trial reported that there was no significant difference in anxiety, depression, adverse events profile, neurotoxicity or global quality of life at 2 years when patients were receiving the newer antiepileptic drugs lamotrigine or topiramate compared with patients receiving valproic acid using results from the NEWQOL. The trial also reported that there was no significant difference in health related quality of life at 2 years when the newer antiepileptic drugs lamotrigine or topiramate were compared with valproic acid using results from the EQ-5D.

One trial reported information regarding HRQoL using the QOLIE-89 inventory when topiramate was compared with valproic acid.73 The trial reported the mean total QOLIE-89 score at baseline and during maintenance treatment for patients receiving topiramate or valproic acid treatment. There was no significant improvement in total QOLIE-89 score when topiramate was compared with valproic acid (weighted mean difference [WMD] -2 [-1 to -3]).

One trial reported information regarding HRQoL using the QOLIE-89 inventory when lamotrigine was compared with valproic acid.71 The trial reported the likelihood that patients would have an improvement in HRQoL when lamotrigine was compared with valproic acid for health perception, energy/fatigue, social isolation, medication effects and attention/concentration subscales of the QOLIE-89 inventory. Patients treated with lamotrigine had a significant increase in the odds of improvement in the health perception (OR 4.0 [1.6 to 10.6]), energy/ fatigue (OR 2.3 [1.1 to 5.3]) and social isolation (OR 2.8 [1.1 to 7.6]) subscales compared with those treated with valproic acid.

Innovator Versus Generic Antiepileptic Drug Evaluation

None of the available controlled clinical trials or observational studies reported health-related quality of life as an endpoint.

Loss of Driver's License or Employment
Innovator Versus Generic Antiepileptic Drug Evaluation

None of the available controlled clinical trials or observational studies reported loss of driver's license or employment as an endpoint.

Seizure Outcomes
Time to First Seizure
Older Versus Newer Antiepileptic Drug Evaluation
Carbamazepine Versus Newer

Six randomized controlled trials reported the time to first seizure for patients when newer antiepileptics were compared with carbamazepine52,55,63,74,77,85 and four trials were amenable for pooling.52,55,63,85

One randomized controlled trial reported the time to first seizure when gabapentin was compared with carbamazepine.85 The time to first seizure was significantly increased when gabapentin was compared with carbamazepine (HR 1.35 [1.14 to 1.60]) (Appendix J Figure 4).

Three randomized controlled trials reported the time to first seizure when lamotrigine was compared with carbamazepine and all three were amenable for pooling.52,55,85 The time to first seizure was nonsignificantly decreased when lamotrigine was compared with carbamazepine (HR 0.97 [0.73 to 1.28]) (Appendix J Figure 4). A high level of statistical heterogeneity was detected (I2: 56.5 percent) but publication bias was not detected (Egger's: p=0.135).

One randomized controlled trial reported the time to first seizure when oxcarbazepine was compared with carbamazepine.85 The time to first seizure was nonsignificantly increased when oxcarbazepine was compared with carbamazepine (HR 1.06 [0.84 to 1.33]) (Appendix J Figure 4).

One randomized controlled trial reported the time to first seizure when topiramate was compared with carbamazepine.85 The time to first seizure was nonsignificantly increased when topiramate was compared with carbamazepine (HR 1.05 [0.89 to 1.25]) (Appendix J Figure 4).

One randomized controlled trial reported the time to first seizure when vigabatrin was compared with carbamazepine.63 The time to first seizure was significantly increased when vigabatrin was compared with carbamazepine (HR 1.79 [1.33 to 2.40]) (Appendix J Figure 4).

Four randomized controlled trials reported the time to first seizure for patients when gabapentin, lamotrigine, oxcarbazepine, topiramate and vigabatrin were compared with carbamazepine and all four trials were amenable for pooling.52,55,60,85 The time to first seizure was nonsignificantly increased when either newer agent was compared with carbamazepine (HR 1.14 [0.98 to 1.33]) (Appendix J Figure 4). A high level of statistical heterogeneity was detected (I2: 66.4 percent) but publication bias was not detected (Egger's: p=0.382).

Two trials were not included in the pooled analysis because the time to first seizure was reported for the whole patient population and not per treatment group but the significance of the inter-group comparison was given. A nonsignificant difference in the time to first seizure when topiramate was compared with carbamazepine was reported in these two trials (Appendix G Table 10).74,77

Phenytoin Versus Newer

Two randomized controlled trials reported time to first seizure when newer antiepileptics were compared with phenytoin and both were amenable for pooling.64,90

One randomized controlled trial reported time to first seizure when lamotrigine was compared with phenytoin.64 The time to first seizure was nonsignificantly increased when lamotrigine was compared with phenytoin (HR 1.40 [0.80 to 2.30]) (Appendix J Figure 5). Since only one trial was available, tests for statistical heterogeneity and publication bias could not be performed.

One randomized controlled trial reported time to first seizure when topiramate was compared with phenytoin.90 The time to first seizure was nonsignificantly increased when topiramate was compared with phenytoin (HR 2.00 [0.98 to 4.12]) (Appendix J Figure 5). Since only one trial was available, tests for statistical heterogeneity and publication bias could not be performed.

Two randomized controlled trials reported time to first seizure when lamotrigine or topiramate were compared with phenytoin and both were amenable for pooling.64,90 Time to first seizure was significantly increased when either newer agent was compared with phenytoin (HR 1.59 [1.04 to 2.43]) (Appendix J Figure 5). Since only two trials were available, tests for statistical heterogeneity and publication bias could not be performed.

Valproic Acid Versus Newer. Three randomized controlled trials reported time to first seizure when newer antiepileptics were compared with valproic acid.74,77,85

One randomized controlled trial reported time to first seizure when lamotrigine was compared with valproic acid.85 The time to first seizure was significantly decreased when lamotrigine was compared with valproic acid (HR 0.71 [0.57 to 0.88]) (Appendix J Figure 6). Since only one trial was available, tests for statistical heterogeneity and publication bias could not be performed.

One randomized controlled trial reported time to first seizure when topiramate was compared with valproic acid.85 The time to first seizure was nonsignificantly decreased when topiramate was compared with valproic acid (HR 0.91 [0.73 to 1.14]) (Appendix J Figure 6). Since only one trial was available, tests for statistical heterogeneity and publication bias could not be performed.

One randomized controlled trial reported the time to first seizure when lamotrigine or topiramate was compared with valproic acid.85 The time to first seizure was nonsignificantly decreased when either newer agent was compared with valproic acid (HR 0.8 [0.63 to 1.02]) (Appendix J Figure 6). Since only one trial was available, tests for statistical heterogeneity and publication bias could not be performed.

Two trials were not included in the pooled analysis because the time to first seizure was reported for the whole patient population and not per treatment group but the significance of the inter-group comparison was given. A nonsignificant difference in the time to first seizure was reported when topiramate was compared with valproic acid in these two trials.74,77

Innovator Versus Generic Antiepileptic Drug Evaluation

None of the available controlled clinical trials or observational studies time to first seizure as an endpoint.

Seizure Occurrence/Breakthrough
Older Versus Newer Antiepileptic Drug Evaluation

None of the available controlled clinical trials or observational studies seizure occurrence or breakthrough as an endpoint.

Innovator Versus Generic Antiepileptic Drug Evaluation

Seven randomized controlled trials reported the occurrence of seizures while patients were receiving innovator versus their associated generic antiepileptic medications109,110,112,114,115,120,123 (Appendix J Figure 7).

Five trials reported data on the occurrence of seizures while patients were receiving innovator versus one or more generic versions of carbamazepine and were all suitable for meta-analysis.109,110,112,114,115 Only one of the trials112 utilized a discernable “A” rated generic carbamazepine product. The risk of experiencing a seizure is nonsignificantly decreased by 14 percent when generic carbamazepine was used versus innovator carbamazepine (RR 0.86 [0.55 to 1.32]). No significant statistical heterogeneity (I2: 0 percent) or publication bias (Egger's: p=0.178) was detected.

One trial reported data on the occurrence of seizures while patients were receiving an innovator versus three generic versions of phenytoin120 but did not use discernable Food and Drug Administration “A” rated generics. The risk of experiencing a seizure is nonsignificantly decreased by 60 percent when generic phenytoin is used versus innovator phenytoin (RR 0.40 [0.14 to 1.12]).

One trial reported data on the occurrence of seizures while patients were receiving innovator versus a generic version of valproic acid123 but did not use discernable FDA “A” rated generics. The risk of experiencing a seizure is nonsignificantly increased by 5 percent when generic valproic acid is used versus innovator valproic acid (RR 1.05 [0.65 to 1.70]).

Seven trials reported data on the occurrence of seizures for any innovator versus generic versions of antiepileptic medication and were all suitable for meta-analysis.109,110,112,114,115,120,123 The risk of experiencing a seizure is nonsignificantly decreased by 13 percent when generic antiepileptic medications are used versus their associated innovator products (RR 0.87 [0.64 to 1.18]). No significant statistical heterogeneity was detected (I2: 0 percent), but publication bias was detected (Egger's: p=0.0004).

The BCS Class I analysis was comprised of the single trial evaluating innovator versus generic versions of valproic acid the results were the same as the aforementioned analysis (RR 1.05 [0.65, 1.70]).

The BCS Class II analysis was comprised of the six trials comparing innovator and generic versions of carbamazepine and phenytoin.109,110,112,114,115,120 The risk of experiencing a seizure is nonsignificantly decreased by 24 percent when generic BCS Class II antiepileptic medications were used versus their associated innovator antiepileptic medications (RR 0.76 [0.51 to 1.14]). No significant statistical heterogeneity was detected (I2: 0 percent), but publication bias was detected (Egger's: p=0.001).

One randomized controlled trial reported data on the occurrence of seizures for an innovator versus a discernable FDA “A” rated generic product, in this case carbamazepine.112 The risk of experiencing a seizure is nonsignificantly decreased by 50 percent when “A” rated generic carbamazepine is used versus innovator carbamazepine (RR 0.50 [0.11 to 2.09]).

Seizure Frequency
Older Versus Newer Antiepileptic Drug Evaluation
Carbamazepine Versus Newer

One randomized controlled trial reported change from baseline in seizure frequency when oxcarbazepine was compared with carbamazepine.49 Seizure frequency was nonsignificantly decreased from baseline when oxcarbazepine was compared with carbamazepine (MD -3 [-6.32 to 0.32]).

One observational study reported the mean percentage change from baseline in seizure frequency when the newer agent topiramate was compared with carbamazepine.83 The study reported that there was no difference between the mean percentage change from baseline in seizure frequency during the maintenance period, but did not provide a measure of deviation.

Phenytoin Versus Newer

Two randomized controlled trials reported change from baseline in seizure frequency when the newer agent oxcarbazepine was compared with phenytoin.57,59 The weighted mean difference for the change in seizure frequency could not be calculated because neither trial reported a measure of deviation for the mean seizure frequency at baseline and both trials reported a different number of patients at baseline and during the maintenance period.

Valproic Acid Versus Newer

One randomized controlled trial reported the change from baseline in seizure frequency when the newer agent oxcarbazepine was compared with valproic acid.58 The weighted mean difference for the change in seizure frequency could not be calculated because the trial did not report a measure of deviation for the mean seizure frequency at baseline and the number of patients at baseline was different than the number of patients during the maintenance period.

Innovator Versus Generic Antiepileptic Drug Evaluation

Three randomized controlled trials reported the occurrence of seizure frequency while patients were receiving innovator and their associated generic antiepileptic medications.109,112,123

Two trials reported data on seizure frequency in patients receiving innovator carbamazepine versus a generic version and were both amenable to pooling.109,112 Only one of the trials utilized a discernable Food and Drug Administration (FDA) “A” rated generic carbamazepine product.112 The seizure frequency is nonsignificantly higher in the generic carbamazepine groups versus innovator carbamazepine group (WMD 0.03 [-0.08 to 0.14] seizures over the evaluative period).

One trial reported data on seizure frequency in patients receiving innovator valproic acid versus a generic version but did not use a discernable FDA “A” rated generic.123 The seizure frequency is nonsignificantly lower in the generic carbamazepine versus innovator carbamazepine group (MD -1.06 [-16.05 to 13.93] seizures over the evaluative period).

Three trials reported data on seizure frequency in patients receiving any innovator versus a generic version of antiepileptic medication and were all suitable for meta-analysis.109,112,123 (Appendix J Figure 8) The seizure frequency is nonsignificantly higher in the generic antiepileptic medication group versus the innovator group (standardized mean difference [SMD] 0.03 [-0.08 to 0.14] seizures over the evaluative period). No statistical heterogeneity was detected (I2: 0 percent), and publication bias could not be evaluated.

The Biopharmaceutics Classification System (BCS) Class I analysis is the same as that reported for the valproic acid analysis (WMD -1.06 [-16.05 to 13.93] seizures over the evaluative period) while the BCS Class II analysis is the same as the carbamazepine analysis (WMD 0.03 [-0.08 to 0.14] seizures over the evaluative period).

One randomized controlled trial reported on the comparison of an innovator to a discernable FDA “A” rated generic product, in this case carbamazepine.112 The seizure frequency is nonsignificantly higher in the generic versus innovator carbamazepine group (MD 0.03 [-0.08 to 0.14] seizures over the evaluative period).

Total Withdrawals
Older Versus Newer Antiepileptic Drug Evaluation
Carbamazepine Versus Newer

Twenty-one studies (16 randomized controlled trials and 5 observational studies) reported withdrawals for any reason while patients were receiving a newer antiepileptic medication compared with carbamazepine.47,52,53,55,60,62,63,70,74,76-80,83,85,98,99,101,103,104

Sixteen randomized controlled trials reported data on withdrawals for any reason when newer antiepileptics were compared versus carbamazepine47,52,53,55,60,62,63,70,74,76-80,83,85 and 14 were all amenable to pooling.52,53,55,60,62,63,70,74,76,78-80,83,85

Three randomized controlled trials reported data on withdrawals for any reason when gabapentin was compared versus carbamazepine and all three were amenable for pooling.60,78,85 The risk of withdrawal was nonsignificantly decreased by 1 percent when gabapentin was compared versus carbamazepine (RR 0.99 [0.83 to 1.17]) ( Appendix J Figure 9). A high level of statistical heterogeneity was detected (I2: 64.1 percent), but publication bias was not detected (Egger's: p=0.739).

Eight randomized controlled trials reported data on withdrawals for any reason when lamotrigine was compared versus carbamazepine and all eight were amenable for pooling.52,55,62,70,76,78,80,85 The risk of withdrawal was significantly decreased by 26 percent when lamotrigine was used versus carbamazepine (RR 0.74 [0.64 to 0.86]) (Appendix J Figure 9). Given the risk difference (RD) (RD -0.101 [-0.158 to -0.0453]), for every 10 patients treated, 1 less patient would withdraw from treatment with lamotrigine than with carbamazepine. Moderate statistical heterogeneity was detected (I2: 48.5 percent), but publication bias was not detected (Egger's: p=0.309). Seven of eight trials had the same direction of effect but differed as to the magnitude of effect.

One randomized controlled trial reported data on withdrawal for any reason when oxcarbazepine was compared versus carbamazepine.85 The risk of withdrawal was nonsignificantly decreased by 7 percent when oxcarbazepine was used versus carbamazepine (RR 0.93 [0.80 to 1.07]) (Appendix J Figure 9).

Three randomized controlled trials reported data on withdrawals for any reason when topiramate was compared versus carbamazepine and were amenable for pooling.74,83,85 The risk of withdrawal was nonsignificantly increased by 6 percent when topiramate was used versus carbamazepine (RR 1.06 [0.96 to 1.17]) (Appendix J Figure 9). No statistical heterogeneity (I2: 0 percent) or publication bias was detected (Egger's: p= 0.641).

Three randomized controlled trials reported data on withdrawals for any reason when vigabatrin was compared versus carbamazepine and were amenable for pooling.53,63,79 The risk of withdrawal was nonsignificantly increased by 3 percent when vigabatrin was used versus carbamazepine (RR 1.03 [0.79 to 1.36]) (Appendix J Figure 9). A low level of statistical heterogeneity was detected (I2: 17.5 percent).

In the pooled analysis of 14 randomized trials reporting data on withdrawals for any reason when either lamotrigine, oxcarbazepine, topiramate, or vigabatrin were compared with carbamazepine.52,53,55,60,62,63,70,74,76,78-80,83,85 The risk of withdrawal was nonsignificantly decreased by 10 percent when all newer antiepileptics were used versus carbamazepine (RR 0.90 [0.82 to 1.00]) (Appendix J Figure 9). A high level of statistical heterogeneity was detected (I2: 55.2 percent), but publication bias was not detected (Egger's: p= 0.689).

Two randomized controlled trials reported data on withdrawals for any reason when the newer antiepileptics gabapentin, lamotrigine, and topiramate were compared with carbamazepine, but these trials were not included in the pooled analysis because withdrawals were reported for the whole patient population and not per treatment group (Appendix G Table 12).47,77

Two observational studies reported withdrawals for any reason when lamotrigine was compared versus carbamazepine.99,101 The risk of withdrawal was significantly decreased by 34 percent when lamotrigine was compared versus carbamazepine (RR 0.66 [0.49 to 0.89]) (Appendix J Figure 10). Given the RD (RD -0.11 [-0.33, 0.12]), for every 10 patients treated with lamotrigine, 1 less patient would withdraw compared with those treated with carbamazepine. Since only two studies were available, tests for statistical heterogeneity and publication bias could not be performed. However, the pooled result was driven almost entirely by one single study, and the effect of the direction differed between the two studies.

One observational study reported withdrawal for any reason when levetiracetam was compared versus carbamazepine.103 The risk of withdrawal was nonsignificantly increased by 67 percent when levetiracetam was compared versus carbamazepine (RR 1.67 [0.73 to 4.35]) (Appendix J Figure 10).

One observational study reported withdrawal for any reason when topiramate was compared versus carbamazepine.104 The risk of withdrawal was nonsignificantly decreased by 11 percent when topiramate was compared versus carbamazepine (RR 0.89 [0.74 to 1.05]) (Appendix J Figure 10).

One observational study reported withdrawal for any reason when vigabatrin was compared versus carbamazepine.98 The risk of withdrawal was significantly increased by 20.5-fold when vigabatrin was compared versus carbamazepine (RR 20.52 [2.22 to 202.53]) (Appendix J Figure 10). Given the RD, (RD 0.24 [0.10 to 0.38]), for every five patients treated with vigabatrin, one additional patient would withdraw compared with those treated with carbamazepine.

Five observational studies reported withdrawals for any reason when lamotrigine, levetiracetam, or vigabatrin was compared with carbamazepine.98,99,101,103,104 The risk of withdrawal was nonsignificantly decreased by 7 percent when either newer agent was compared versus carbamazepine (RR 0.93 [0.61 to 1.40]) (Appendix J Figure 10). A high level of statistical heterogeneity was detected (I2: 60.9 percent), but no significant publication bias was detected (Egger's: p=0.307).

Two randomized controlled trials reported withdrawals for any reason when newer antiepileptics were compared with controlled or sustained release carbamazepine and both were amenable for pooling.81,86

One randomized controlled trial reported withdrawals for any reason when lamotrigine was compared with sustained release carbamazepine.86 The risk of withdrawal was nonsignificantly decreased by 18 percent when lamotrigine was compared versus sustained release carbamazepine (RR 0.82 [0.52 to 1.27]) (Appendix J Figure 11).

One randomized controlled trial reported withdrawals for any reason when levetiracetam was compared with controlled release carbamazepine.81 There was no significant difference in the risk of withdrawal when levetiracetam was compared versus controlled release carbamazepine in patients (RR 1.00 [0.79 to 1.26]) (Appendix J Figure 11).

Two randomized controlled trials reported withdrawals for any reason when lamotrigine or levetiracetam were compared with controlled or sustained release carbamazepine and both were amenable for pooling.81,86 The risk of withdrawal was nonsignificantly decreased by 4 percent when either newer agent was compared with controlled or sustained release carbamazepine (RR 0.96 [0.78 to 1.18]) (Appendix J Figure 11).

Phenytoin Versus Newer

Three randomized controlled trials reported withdrawals for any reason while patients were receiving a newer antiepileptic medication compared with phenytoin and all were amenable for pooling.57,59,64

One randomized controlled trial reported data on withdrawals for any reason when lamotrigine was compared with phenytoin.64 The risk of withdrawal was nonsignificantly decreased by 1 percent when lamotrigine was compared versus phenytoin (RR 0.99 [0.75 to 1.31]) (Appendix J Figure 12).

Two randomized controlled trials reported data on withdrawals for any reason when oxcarbazepine was compared with phenytoin and were amenable for pooling.57,59 The risk of withdrawal was nonsignificantly decreased by 15 percent when oxcarbazepine was compared versus phenytoin (RR 0.85 [0.66 to 1.09]) (Appendix J Figure 12).

In the pooled analysis of three randomized trials reporting data on withdrawals for any reason either lamotrigine or oxcarbazepine were compared versus phenytoin.57,59 The risk of withdrawal was nonsignificantly decreased by 9 percent when the newer agents were compared versus phenytoin (RR 0.91 [0.76 to 1.09]) (Appendix J Figure 12). No significant statistical heterogeneity was detected (I2: 0 percent), however tests for publication bias could not be performed.

Valproic Acid Versus Newer

Nineteen studies (17 randomized controlled trials and 2 observational studies) reported withdrawals for any reason while patients were receiving a newer antiepileptic medication compared with valproic acid.50,58,61,65,68,72-77,80,85,87-89,94,99,101

Seventeen randomized controlled trials reported data on withdrawals for any reason when newer antiepileptics were compared versus valproic acid50,58,61,65,68,72-77,80,85,87-89,94 and 16 were amenable for pooling.50,58,61,65,68,72-76,80,85,87-89,94

One randomized controlled trial reported data on withdrawals for any reason when felbamate was compared versus valproic acid.50 There was no significant difference in the risk of overall withdrawal when felbamate was compared versus valproic acid (RR 1.00 [0.11 to 9.24]) (Appendix J Figure 13). Since only one trial was available, tests for statistical heterogeneity and publication bias could not be performed.

Ten randomized controlled trials reported data on withdrawals for any reason when lamotrigine was compared versus valproic acid and were amenable for pooling.68,72,75,76,80,85,87-89,94 The risk of withdrawal was significantly decreased by 12 percent when lamotrigine was compared versus valproic acid (RR 0.88 [0.78 to 0.99]) (Appendix J Figure 13). No statistical heterogeneity (I2: 0 percent) or publication bias was detected (Egger's: p=0.953).

One randomized controlled trial reported data on withdrawals for any reason when oxcarbazepine was compared versus valproic acid.58 The risk of withdrawal was nonsignificantly increased by 20 percent when oxcarbazepine was compared versus valproic acid (RR 1.20 [0.87 to 1.66]) (Appendix J Figure 13). Since only one trial was available, tests for statistical heterogeneity or publication bias could not be performed.

Four randomized controlled trials reported data on withdrawals for any reason when topiramate was compared versus valproic acid and were amenable for pooling.65,73,74,85 The risk of withdrawal was nonsignificantly decreased by 2 percent when topiramate was compared versus valproic acid (RR 0.98 [0.80 to 1.21]) (Appendix J Figure 13). A moderate level of statistical heterogeneity was detected (I2: 38.5 percent), but publication bias was not detected (Egger's: p=0.922).

One randomized controlled trial reported data on withdrawals for any reason when vigabatrin was compared versus valproic acid.61 The risk of withdrawal was nonsignificantly decreased by 32 percent when vigabatrin was compared versus valproic acid (RR 0.68 [0.42 to 1.08]) (Appendix J Figure 13). Since only one trial was available, tests for statistical heterogeneity and publication bias could not be performed.

In the pooled analysis of 16 randomized controlled trials reporting data on withdrawals for any reason when felbamate, lamotrigine, oxcarbazepine, topiramate or vigabatrin was compared versus valproic acid.50,58,61,65,68,72-76,80,85,87-89,94 The risk of withdrawal was nonsignificantly decreased by 4 percent when the newer agents were compared versus valproic acid (RR 0.96 [0.85 to 1.09]) (Appendix J Figure 13). No statistical heterogeneity (I2: 0 percent) or publication bias (Egger's: p=0.959) was detected.

One randomized controlled trial reported data on withdrawals for any reason when topiramate was compared with valproic acid, but this trial was not included in the pooled analysis because withdrawals were reported for the whole patient population and not per treatment group (Appendix G Table 12).77

Two observational studies reported withdrawals for any reason while patients were receiving newer antiepileptic medications compared with valproic acid and both were amenable for pooling.99,101

Two observational studies reported withdrawals for any reason when lamotrigine was compared with valproic acid and were amenable for pooling.99,101 The risk of withdrawal was nonsignificantly decreased by 9 percent when lamotrigine were compared versus valproic acid (RR 0.91 [0.63 to 1.30]) (Appendix J Figure 14). Since only two studies were available, tests for statistical heterogeneity and publication bias could not be performed.

Ethosuximide Versus Newer

One randomized controlled trial reported withdrawals for any reason while patients were receiving lamotrigine compared with ethosuximide.89 The risk of overall withdrawal was nonsignificantly decreased by 5 percent when lamotrigine was compared versus ethosuximide (RR 0.95 [0.53 to 1.71]). Since only one trial was available, tests for statistical heterogeneity and publication bias could not be performed.

Innovator Versus Generic Antiepileptic Drug Evaluation

Eleven studies (nine randomized controlled trials, one prospective nonrandomized trial, and one observational study) reported withdrawals for any reason while patients were receiving innovator and their associated generic antiepileptic medications.108-113,115,117,120,121,130

Seven randomized controlled trials reported data on withdrawals for any reason for innovator versus one or more generic versions of carbamazepine108-113,115 and were all amenable to pooling. Only one trial112 used discernable FDA “A” rated generics. The risk of withdrawals for any reason is nonsignificantly decreased by 15 percent when generic carbamazepine was used versus innovator carbamazepine (RR 0.85 [0.30 to 2.40]) (Appendix J Figure 15). No significant statistical heterogeneity (I2: 0 percent) or publication bias was detected (Egger's: p=0.42). One observational study compared innovator carbamazepine to a generic version.130 The risk of withdrawals for any reason is significantly increased by 17 percent when generic carbamazepine was used versus innovator carbamazepine (RR 1.17 [1.04 to 1.31]) (Appendix J Figure 15). Tests for statistical heterogeneity or publication bias could not be calculated.

Three controlled clinical trials (two randomized controlled trials,120,121 and one prospective nonrandomized trial) 117 evaluated innovator phenytoin versus generic phenytoin and were suitable for meta-analysis. None of the trials used discernable FDA “A” rated generics. The risk of withdrawals for any reason was equivalent when generic phenytoin was used versus innovator phenytoin (RR 1.00 [0.24 to 4.14]) (Appendix J Figure 15). No significant statistical heterogeneity was detected (I2: 0 percent), but publication bias was detected (Egger's: p=0.04).

Ten trials (nine randomized controlled trials,108-113,115,120,121 and one prospective nonrandomized trial)117 reported data on withdrawals for any reason for any innovator versus a generic version of antiepileptic medication and were all suitable for meta-analysis. The risk of withdrawals for any reason is nonsignificantly decreased by 10 percent when generic antiepileptic medications were used versus their appropriate innovator antiepileptic medications (RR 0.90 [0.39, 2.08]) (Appendix J Figure 15). No significant statistical heterogeneity (I2: 0 percent) or publication bias was detected (Egger's: p=0.45). Analysis on any innovator versus generic antiepileptic medication for observational trials is the same as that reported for the carbamazepine observational trial analysis (RR 1.17 [1.04 to 1.31]) (Appendix J Figure 15).

None of the available trials or studies evaluated BCS Class I antiepileptic medications, so analysis was not possible. The BCS Class II analysis for controlled trials is the same as the “any antiepileptic medications” controlled trial analysis (RR 0.90 [0.39 to 2.08]) and the BCS Class II analysis for observational trials is the same as carbamazepine observational trial analysis (RR 1.17 [1.04 to 1.31]) (Appendix J Figure 15).

One randomized controlled trial reported on the comparison of an innovator to a discernable FDA “A” rated generic product, in this case carbamazepine.112 The risk of withdrawals for any reason is nonsignificantly decreased by 20 percent when “A” rated generic carbamazepine was used versus innovator carbamazepine (RR 0.80 [0.12 to 5.16]) (Appendix J Figure 15). Tests for statistical heterogeneity or publication bias could not be calculated.

Withdrawals Due to Lack of Efficacy
Older Versus Newer Antiepileptic Drug Evaluation
Carbamazepine Versus Newer

Sixteen studies (11 randomized controlled trials and 5 observational studies) reported withdrawals due to lack of efficacy while patients were receiving a newer antiepileptic medication compared with carbamazepine.49,52,53,55,62,63,74,77,78,80,85,98,99,101,104,105

Eleven randomized controlled trials reported data on withdrawals due to lack of efficacy when newer antiepileptics were compared versus valproic acid.49,52,53,55,62,63,74,77,78,80,85 and 10 were amenable for pooling.49,52,53,55,62,63,74,78,80,85

Two randomized controlled trials reported data on withdrawals due to lack of efficacy when gabapentin was compared versus carbamazepine and were amenable for pooling.78,85 The risk of withdrawal was significantly increased by 2.3-fold when gabapentin was compared versus carbamazepine (RR 2.25 [1.64 to 3.08]) (Appendix J Figure 16). Given the RD (RD 0.08 [-0.09 to 0.25]), for every 13 patients treated with gabapentin, 1 additional patient would withdraw due to lack of effective treatment compared with those treated with carbamazepine. Since only two trials were available, tests for statistical heterogeneity and publication bias could not be performed.

Six randomized controlled trials reported data on withdrawals due to lack of efficacy when lamotrigine was compared versus carbamazepine and were amenable for pooling.52,55,62,78,80,85 The risk of withdrawal due to lack of efficacy was significantly increased by 43 percent when lamotrigine was compared versus carbamazepine (RR 1.43 [1.03 to 1.99]) (Appendix J Figure 16). Given the RD, (RD 0.023 [0.0025 to 0.044]), for every 44 patients treated with lamotrigine, 1 additional patient would withdraw due to lack of efficacy compared with those treated with carbamazepine. No statistical heterogeneity (I2: 0 percent) or publication bias was detected (Egger's: p=0.759).

Two randomized controlled trials reported data on withdrawals due to lack of efficacy when oxcarbazepine was compared versus carbamazepine and were amenable for pooling.49,85 The risk of withdrawal due to lack of effective treatment was nonsignificantly increased by 1 percent when oxcarbazepine was compared versus carbamazepine (RR 1.01 [0.64 to 1.59]) (Appendix J Figure 16). Since only two trials were available, tests for statistical heterogeneity and publication bias could not be performed.

Two randomized controlled trials reported data on withdrawals due to lack of efficacy when topiramate was compared versus carbamazepine and were amenable for pooling.74,85 The risk of withdrawal due to lack of efficacy was nonsignificantly increased by 28 percent when topiramate was compared versus carbamazepine (RR 1.28 [0.93 to 1.78]) (Appendix J Figure 16). No statistical heterogeneity was detected (I2: 0 percent), but tests for publication bias could not be performed.

Two randomized controlled trials reported data on withdrawals due to lack of efficacy when vigabatrin was compared versus carbamazepine and were amenable for pooling.53,63 The risk of withdrawal due to lack of efficacy was significantly increased by 3.0-fold when vigabatrin was compared versus carbamazepine (RR 2.98 [1.58 to 5.61]) (Appendix J Figure 16). Given the RD (RD 0.14 [0.02 to 0.26]), for every eight patients treated with vigabatrin, one additional patient would withdraw due to lack of efficacy compared with those treated with carbamazepine. Since only two trials were available, tests for statistical heterogeneity and publication bias could not be performed.

Ten randomized controlled trials reported data on withdrawals due to lack of efficacy when gabapentin, lamotrigine, oxcarbazepine, topiramate, or vigabatrin was compared versus carbamazepine and were amenable for pooling.49,52,53,55,62,63,74,78,80,85 The risk of withdrawal due to lack of efficacy was significantly increased by 59 percent when newer agents were compared versus carbamazepine (RR 1.59 [1.25 to 2.02]) (Appendix J Figure 16). Given the RD (RD 0.02 [0.003 to 0.04]), for every 50 patients treated with either newer agent, 1 additional patient would withdraw due to lack of efficacy compared with those treated with carbamazepine. No statistical heterogeneity (I2: 0 percent) or publication bias was detected (Egger's: p=0.747).

One randomized controlled trial reported data on withdrawals due to lack of efficacy when topiramate was compared versus carbamazepine.77 The data from this trial was not included in the pooled analysis because withdrawals due to lack of efficacy were reported for the whole population and not per treatment group (Appendix J Figure 16).

Five observational studies reported withdrawals due to lack of efficacy while patients were receiving a newer antiepileptic medications compared with carbamazepine and all five were amenable for pooling.98,99,101,104,105

Two observational studies reported withdrawals due to lack of efficacy when lamotrigine was compared versus carbamazepine and were amenable for pooling.99,101 The risk of withdrawal due to lack of efficacy was nonsignificantly increased by 3 percent when lamotrigine was compared versus carbamazepine (RR 1.03 [0.66 to 1.60]) (Appendix J Figure 17). Since only two trials were available, tests for statistical heterogeneity or publication bias could not be performed.

Two observational studies reported withdrawal due to lack of efficacy when topiramate was compared versus carbamazepine and were amenable for pooling.104,105 The risk of withdrawal due to lack of efficacy was nonsignificantly decreased by 19 percent when topiramate was compared versus carbamazepine (RR 0.81 [0.45 to 1.45]) (Appendix J Figure 17). Since only two trials were available, tests for statistical heterogeneity or publication bias could not be performed.

One observational study reported withdrawal due to lack of efficacy when vigabatrin was compared versus carbamazepine.98 The risk of withdrawal due to lack of efficacy was significantly increased by 18.4-fold when vigabatrin was compared versus carbamazepine (RR 18.36 [1.97 to 182.09]) (Appendix J Figure 17). Given the RD (RD 0.21 [0.08 to 0.35]), for every five patients treated with vigabatrin, one additional patient would withdraw due to lack of efficacy compared with those treated with carbamazepine. Since only one trial was available, tests for statistical heterogeneity or publication bias could not be performed.

Five observational studies reported withdrawals due to lack of efficacy when lamotrigine, topiramate, or vigabatrin was compared with carbamazepine.98,99,101,104,105 The risk of withdrawal due to lack of efficacy was nonsignificantly decreased by 3 percent when newer agents were compared versus carbamazepine (RR 0.97 [0.60 to 1.56]) (Appendix J Figure 17). A lower level of statistical heterogeneity was detected (I2: 35.3 percent), but publication bias was not detected (Egger's: p=0.522).

One randomized controlled trial reported withdrawals due to lack of efficacy when levetiracetam was compared with controlled release carbamazepine.81 The risk of withdrawal due to lack of efficacy was significantly increased by 2.4-fold when levetiracetam was compared versus controlled release carbamazepine (RR 2.43 [1.32 to 4.52]). Given the RD, (RD 0.064 [0.021 to 0.11]), for every 16 patients treated with levetiracetam, 1 additional patient would withdraw due to lack of efficacy compared with those treated with carbamazepine. Since only one trial was available, tests for statistical heterogeneity and publication bias could not be performed.

Phenytoin Versus Newer

Three randomized controlled trials reported withdrawals due to lack of efficacy while patients were receiving newer antiepileptic medications versus phenytoin and all three were amenable for pooling.57,59,64

One randomized controlled trial reported data on withdrawal due to lack of efficacy when lamotrigine was compared versus phenytoin.64 The risk of withdrawal due to lack of efficacy was nonsignificantly decreased by 45 percent when lamotrigine was compared versus phenytoin (RR 0.55 [0.07 to 4.15]) (Appendix J Figure 18). Since only one trial was available, tests for statistical heterogeneity and publication bias could not be performed.

Two randomized controlled trials reported data on withdrawal due to lack of efficacy when oxcarbazepine was compared versus phenytoin and were amenable for pooling.57,59 The risk of withdrawal due to lack of efficacy was nonsignificantly increased by 24 percent when oxcarbazepine was compared versus phenytoin (RR 1.24 [0.34 to 4.55]) (Appendix J Figure 18). Since only two trials were available, tests for statistical heterogeneity and publication bias could not be performed.

In the pooled analysis of three randomized controlled trials reporting data on withdrawals due to lack of efficacy, either lamotrigine or oxcarbazepine were compared versus phenytoin.57,59,64 The risk of withdrawal due to lack of efficacy was nonsignificantly increased by 3 percent when either newer agent was compared versus phenytoin (RR 1.03 [0.33 to 3.23]) (Appendix J Figure 18). No statistical heterogeneity was detected (I2: 0 percent), but tests for publication bias could not be performed.

Valproic Acid Versus Newer

Fifteen studies (12 randomized controlled trials and 3 observational studies) reported withdrawals due to lack of efficacy while patients were receiving a newer antiepileptic medication compared with valproic acid.58,61,65,74,75,77,80,84,85,87,88,94,99,101,105

Twelve randomized controlled trials reported data on withdrawals due lack of efficacy when newer antiepileptics were compared versus valproic acid58,61,65,74,75,77,80,84,85,87,88,94 and 11 were amenable for pooling.58,61,65,74,75,80,84,85,87,88,94

Six randomized controlled trials reported data on withdrawals due to lack of efficacy when lamotrigine was compared versus valproic acid and were amenable for pooling.75,80,85,87,88,94 The risk of withdrawal due to lack of efficacy was nonsignificantly increased by 53 percent when lamotrigine was compared versus valproic acid (RR 1.53 [0.74 to 3.17]) (Appendix J Figure 19). A higher level of statistical heterogeneity was detected (I2: 52.2 percent), but publication bias was not detected (Egger's: p=0.724).

One randomized controlled trial reported data on withdrawals due to lack of efficacy when oxcarbazepine was compared versus valproic acid.58 The risk of withdrawal due to lack of efficacy was nonsignificantly decreased by 5 percent when oxcarbazepine was compared versus valproic acid (RR 0.95 [0.33 to 2.71]) (Appendix J Figure 19). Since only one trial was available, tests for statistical heterogeneity and publication bias could not be performed.

Four randomized controlled trials reported data on withdrawals due to lack of efficacy when topiramate was compared versus valproic acid and were amenable for pooling.65,74,84,85 The risk of withdrawal due to lack of efficacy was nonsignificantly increased by 12 percent when topiramate was compared versus valproic acid (RR 1.12 [0.74 to 1.70]) (Appendix J Figure 19). No statistical heterogeneity (I2: 0 percent) or publication bias was detected (Egger's: p=0.487).

One randomized controlled trial reported data on withdrawals due to lack of efficacy when vigabatrin was compared versus valproic acid.61 The risk of withdrawal due lack of efficacy was nonsignificantly decreased by 43 percent when vigabatrin was compared versus valproic acid (RR 0.57 [0.29 to 1.13]) (Appendix J Figure 19). Since only one trial was available, tests for statistical heterogeneity and publication bias could not be performed.

Eleven randomized controlled trials reported data on withdrawals due to lack of efficacy when either lamotrigine, oxcarbazepine, topiramate or vigabatrin were compared versus valproic acid and were amenable for pooling.28,58,61,65,74,75,80,84,87,88,94 The risk of withdrawal was nonsignificantly increased by 10 percent when all newer agents were compared versus valproic acid (RR 1.10 [0.77 to 1.56]) (Appendix J Figure 19). A lower level of statistical heterogeneity was detected (I2: 26.6 percent), but publication bias was not detected (Egger's: p=0.982).

One randomized controlled trial reported data on withdrawals due to lack of efficacy when topiramate was compared versus valproic acid, but this trial was not included in the pooled analysis because withdrawals were reported for the whole patient population and not per treatment group (Appendix G Table 10).77

Three observational studies reported withdrawals due to lack of efficacy when lamotrigine or topiramate were compared versus valproic acid and all three were amenable for pooling.99,101,105

Two observational studies reported data on withdrawals due to lack of efficacy when lamotrigine was compared versus valproic acid and were amenable for pooling.99,101 The risk of withdrawal due to lack of efficacy was nonsignificantly increased by 1 percent when lamotrigine was compared versus valproic acid (RR 1.01 [0.61 to 1.66]) (Appendix J Figure 20). Since only two trials were available, tests for statistical heterogeneity and publication bias could not be performed.

One observational study reported data on withdrawals due to lack of efficacy when topiramate was compared versus valproic acid.105 The risk of withdrawal due to lack of efficacy was nonsignificantly increased by 2 percent when topiramate was compared versus valproic acid (RR 1.02 [0.65 to 1.60]) (Appendix J Figure 20). Since only one trial was available, tests for statistical heterogeneity and publication bias could not be performed.

Three observational studies reported data on withdrawals due to lack of efficacy when either lamotrigine or topiramate were compared versus valproic acid and were amenable for pooling.99,101,105 The risk of withdrawal due to lack of efficacy was nonsignificantly increased by 2 percent when newer agents were compared versus valproic acid (RR 1.02 [0.73 to 1.42]) (Appendix J Figure 20). No statistical heterogeneity was detected (I2: 0 percent), but tests for publication bias could not be performed.

Innovator Versus Generic Antiepileptic Drug Evaluation

Ten trials (nine randomized controlled trials108-113,115,120,121 and one prospective nonrandomized trial)117 reported withdrawals due to lack of efficacy while patients were receiving innovator and their associated generic antiepileptic medications.

Seven randomized controlled trials reported data on withdrawals due to lack of efficacy for innovator carbamazepine versus one or more generic versions108-113,115 and they were all suitable for meta-analysis. Only one of the trials utilized a discernable FDA “A” rated generic carbamazepine product.112 The risk of withdrawals due to ineffective treatment is nonsignificantly decreased by 13 percent when generic carbamazepine was used versus innovator carbamazepine (RR 0.87 [0.29 to 2.63]) (Appendix J Figure 21). No significant statistical heterogeneity (I2: 0 percent) or publication bias was detected (Egger's: p=0.99).

Three trials (two randomized controlled trials120,121 and one prospective nonrandomized trial117 reported withdrawals due to lack of efficacy while patients were receiving innovator phenytoin versus three generic versions and were all suitable for meta-analysis but the trials did not use discernable FDA “A” rated generics. The risk of withdrawals due to ineffective treatment is nonsignificantly increased by 45 percent when generic phenytoin was used versus innovator phenytoin (RR 1.45 [0.28 to 7.53]) (Appendix J Figure 21). No significant statistical heterogeneity was detected (I2: 0 percent), but publication bias was detected (Egger's: p=0.03).

Ten trials (nine randomized controlled trials108-113,115,120,121 and one prospective nonrandomized trial117) reported withdrawals due to lack of efficacy for any innovator versus generic version of antiepileptic medication and were all suitable for meta-analysis. The risk of withdrawals due to ineffective treatment is nonsignificantly increased by 2 percent when generic antiepileptic medications were used versus their appropriate innovator antiepileptic medications (RR 1.02 [0.41 to 2.54]) (Appendix J Figure 21). No significant statistical heterogeneity (I2: 0 percent) or publication bias was detected (Egger's: p=0.434).

None of the available drugs were in BCS Class I, so analysis was not possible. The BCS Class II analysis is the same as the analysis for any antiepileptic medications (RR 1.02 [0.41 to 2.54]) (Appendix J Figure 21).

One randomized controlled trial reported on the comparison of an innovator to a discernable FDA “A” rated generic product, in this case carbamazepine.112 The risk of withdrawals due to ineffective treatment is nonsignificantly decreased by 25 percent when “A” rated carbamazepine generic was used versus innovator carbamazepine (RR 0.75 [0.15 to 3.73]) (Appendix J Figure 21). Tests for statistical heterogeneity or publication bias could not be calculated.

Seizure Remission
Older Versus Newer Antiepileptic Drug Evaluation
Carbamazepine Versus Newer

Two randomized controlled trials reported the number of patients who achieved 12-month seizure remission when newer antiepileptics were compared with carbamazepine.28,63

One randomized controlled trial reported the number of patients who achieved 12-month seizure remission when gabapentin was compared with carbamazepine.28 The risk of achieving 12-month seizure remission was significantly decreased by 15 percent when gabapentin was compared with carbamazepine (RR 0.85 [0.76 to 0.95]) (Appendix J Figure 22). Given the RD (RD -0.10 [-0.17 to -0.03]), for every 10 patients treated, 1 less patient would achieve 12-month seizure remission when treated with gabapentin versus carbamazepine.

One randomized controlled trial reported the number of patients who achieved 12-month seizure remission when lamotrigine was compared with carbamazepine.28 The risk of achieving 12-month seizure remission was nonsignificantly decreased by 4 percent when lamotrigine was compared with carbamazepine (RR 0.96 [0.87 to 1.06]) (Appendix J Figure 22).

One randomized controlled trial reported the number of patients who achieved 12-month seizure remission when oxcarbazepine was compared with carbamazepine.28 The risk of achieving 12-month seizure remission was significantly decreased by 52 percent when oxcarbazepine was compared with carbamazepine (RR 0.48 [0.39 to 0.59]) (Appendix J Figure 22). Given the RD ([RD -0.36 [-0.44 to -0.28]), for every three patients treated, one less patient would achieve 12-month seizure remission when treated with oxcarbazepine versus carbamazepine.

One randomized controlled trial reported the number of patients who achieved 12-month seizure remission when topiramate was compared with carbamazepine.28 The risk of achieving 12-month seizure remission was significantly decreased by 10 percent when topiramate was compared with carbamazepine (RR 0.90 [0.81 to 0.99]) (Appendix J Figure 22). Given the RD (RD -0.073 [-0.142 to -0.004]), for every 14 patients treated, 1 less patient would achieve 12 month seizure remission when treated with topiramate versus carbamazepine.

One randomized controlled trial reported the number of patients who achieved 12-month seizure remission when vigabatrin was compared with carbamazepine.63 The risk of achieving 12-month seizure remission was nonsignificantly decreased by 5 percent when vigabatrin was compared with carbamazepine (RR 0.95 [0.79 to 1.14]) (Appendix J Figure 22).

Two randomized controlled trials reported the number of patients who achieved 12-month seizure remission when gabapentin, lamotrigine, oxcarbazepine, topiramate or vigabatrin was compared with carbamazepine and were amenable for pooling.28,63 The risk of achieving 12-month seizure remission was significantly decreased by 19 percent when newer agents were compared with carbamazepine (RR 0.81 [0.67 to 0.99]) ( Appendix J Figure 22). Given the RD (RD -0.12 [-0.23 to -0.001]), for every nine patients treated, one less patient would achieve 12-month seizure remission when treated with either newer agent versus carbamazepine. A high level of statistical heterogeneity was detected (I2: 84.7 percent), but publication bias was not detected (Egger's: p=0.069).

One randomized controlled trial reported the number of patients who achieved 24-month seizure remission when newer antiepileptics were compared with carbamazepine.28 This randomized controlled trial reported the number of patients who achieved 24-month seizure remission when gabapentin was compared with carbamazepine.28 The risk of achieving 24-month seizure remission was significantly decreased by 21 percent when gabapentin was compared with carbamazepine (RR 0.79 [0.66 to 0.94]) (Appendix J Figure 23). Given the RD (RD -0.10 [-0.17 to -0.02]), for every 10 patients treated, 1 less patient would achieve 12-month seizure remission when treated with gabapentin versus carbamazepine. Since only one trial was available, tests for statistical heterogeneity and publication bias could not be performed.

One randomized controlled trial reported the number of patients who achieved 24-month seizure remission when lamotrigine was compared with carbamazepine.28 The risk of achieving 24-month seizure remission was nonsignificantly decreased by 8 percent when lamotrigine was compared with carbamazepine (RR 0.92 [0.78 to 1.08]) (Appendix J Figure 23). Since only one trial was available, tests for statistical heterogeneity and publication bias could not be performed.

One randomized controlled trial reported the number of patients who achieved 24-month seizure remission when oxcarbazepine was compared with carbamazepine.28 The risk of achieving 24-month seizure remission was significantly decreased by 27 percent when oxcarbazepine was compared with carbamazepine (RR 0.73 [0.58 to 0.91]) (Appendix J Figure 23). Given the RD (RD -0.12 [-0.21 to -0.04]), for every nine patients treated, one less patient would achieve 12-month seizure remission when treated with oxcarbazepine versus carbamazepine. Since only one trial was available, tests for statistical heterogeneity and publication bias could not be performed.

One randomized controlled trial reported the number of patients who achieved 24-month seizure remission when topiramate was compared with carbamazepine.28 The risk of achieving 24-month seizure remission was significantly decreased by 16 percent when topiramate was compared with carbamazepine (RR 0.84 [0.71 to 0.94]) (Appendix J Figure 23). Given the RD (RD -0.07 [-0.15 to -0.001]), for every 15 patients treated,1 less patient would achieve 12-month seizure remission when treated with topiramate versus carbamazepine. Since only one trial was available, tests for statistical heterogeneity and publication bias could not be performed.

One randomized controlled trial reported the number of patients who achieved 24-month seizure remission when gabapentin, lamotrigine, oxcarbazepine, or topiramate was compared with carbamazepine.28 The risk of achieving 24-month seizure remission was significantly decreased by 18 percent when these newer antiepileptic medications was compared with carbamazepine (RR 0.82 [0.72 to 0.94]) (Appendix J Figure 23). Given the RD (RD -0.08 [-0.14 to -0.02]), for every 13 patients treated with a newer antiepileptic medication, 1 less patient would achieve 12-month seizure remission when treated with a newer agent versus carbamazepine. Since only one trial was available, tests for statistical heterogeneity and publication bias could not be performed.

Valproic Acid Versus Newer

One randomized controlled trial reported the number of patients who achieved 12-month seizure remission when newer antiepileptics were compared with valproic acid.29

One randomized controlled trial reported the number of patients who achieved 12-month seizure remission when lamotrigine was compared with valproic acid.29 The risk of achieving 12-month seizure remission was nonsignificantly decreased by 6 percent when lamotrigine was compared with valproic acid (RR 0.94 [0.84 to 1.04]) (Appendix J Figure 24). Since only one trial was available, tests for statistical heterogeneity and publication bias could not be performed.

One randomized controlled trial reported the number of patients who achieved 12-month seizure remission when topiramate was compared with valproic acid.29 The risk of achieving 12-month seizure remission remained the same when topiramate was compared with valproic acid (RR 1.00 [0.90 to 1.10]) (Appendix J Figure 24).

One randomized controlled trial reported the number of patients who achieved 12-month seizure remission when lamotrigine or topiramate was compared with valproic acid.29 The risk of achieving 12-month seizure remission was nonsignificantly decreased when either newer agent was compared with valproic acid (RR 0.97 [0.89 to 1.06]) (Appendix J Figure 24).

One randomized controlled trial reported the number of patients who achieved 24-month seizure remission when newer antiepileptics were compared with valproic acid.85

One randomized controlled trial reported the number of patients who achieved 24-month seizure remission when lamotrigine was compared with valproic acid.85 The risk of achieving 24-month seizure remission was nonsignificantly decreased by 17 percent when lamotrigine was compared with valproic acid (RR 0.83 [0.68 to 1.00]) (Appendix J Figure 25).

One randomized controlled trial reported the number of patients who achieved 24-month seizure remission when topiramate was compared with valproic acid.85 The risk of achieving 24-month seizure remission was nonsignificantly decreased by 12 percent when topiramate was compared with valproic acid (RR 0.88 [0.73 to 1.05]) (Appendix J Figure 25).

One randomized controlled trial reported the number of patients who achieved 24-month seizure remission when lamotrigine or topiramate was compared with valproic acid.85 The risk of achieving 24-month seizure remission was nonsignificantly decreased by 15 percent when either newer antiepileptic was compared with valproic acid (RR 0.85 [0.73 to 1.00]) (Appendix J Figure 25).

Innovator Versus Generic Antiepileptic Drug Evaluation

There were no controlled clinical trials or controlled observational studies that reported data on seizure remission.

Seizure Freedom for Study Duration
Older Versus Newer Antiepileptic Drug Evaluation
Carbamazepine Versus Newer

Twenty-one studies (15 randomized controlled trials and 6 observational studies) reported data on seizure freedom for the duration of the study while receiving a newer antiepileptic medications compared with carbamazepine.48,49,52-56,62,70,74,76-78,80,82,83,98,99,103,105,164

Fifteen randomized controlled trials reported data on seizure freedom for the duration of the study when newer antiepileptics were compared versus carbamazepine48,49,52,53,55,56,62,70,74,76-78,80,82,83 and were all amenable to pooling.

One randomized controlled trial comparing gabapentin to carbamazepine reported data on seizure freedom for the study duration.78 The risk of remaining seizure free for the duration of the study is significantly decreased by 26 percent when gabapentin is used versus carbamazepine (RR 0.74 [0.56 to 0.97]) (Appendix J Figure 26).

Seven randomized controlled trials comparing lamotrigine to carbamazepine reported data on seizure freedom for the study duration.52,55,62,70,76,78,80 The risk of remaining seizure free for the duration of study is nonsignificantly decreased by 6 percent when lamotrigine is used versus carbamazepine (RR 0.94 [0.83 to 1.05]) (Appendix J Figure 26). A low level of statistical heterogeneity was detected (I2: 35.5 percent), but no significant publication bias was detected (Egger's p=0.073).

Three trials comparing oxcarbazepine to carbamazepine reported data on seizure freedom for the study duration.48,49,82 The risk of remaining seizure free for the duration of study is nonsignificantly decreased by 4 percent when oxcarbazepine is used versus carbamazepine (RR 0.96 [0.77 to 1.19]) (Appendix J Figure 26). No statistical heterogeneity was detected (I2: 0 percent).

Three trials comparing topiramate to carbamazepine reported data on seizure freedom for the study duration.74,77,83 The risk of remaining seizure free for the duration of study is nonsignificantly increased by 8 percent when topiramate is used versus carbamazepine (RR 1.08 [0.91 to 1.27]) (Appendix J Figure 26). No significant statistical heterogeneity was detected (I2: 0 percent), but publication bias was detected (Egger's p=0.021).

Two trials comparing vigabatrin to carbamazepine reported data on seizure freedom for the study duration.53,56 The risk of remaining seizure free for the duration of study is nonsignificantly decreased by 30 percent when vigabatrin is used versus carbamazepine (RR 0.70 [0.49 to 1.01]) (Appendix J Figure 26).

Fifteen randomized controlled trials comparing newer antiepileptic medications to carbamazepine reported data on seizure freedom for the study duration.48,49,52,53,55,56,62,70,74,76-78,80,82,83 The risk of remaining seizure free for the duration of study is nonsignificantly decreased by 6 percent when newer antiepileptic drugs (AEDs) are used versus carbamazepine (RR 0.94 [0.87 to 1.03]) (Appendix J Figure 26). A low level of statistical heterogeneity was detected (I2: 21.5 percent), and a trend toward publication bias was detected (Egger's p=0.054).

Six observational studies reported seizure freedom for study duration while patients were receiving newer antiepileptic medications compared with carbamazepine and all six were amenable for pooling.54,98,99,103,105,164

One observational study comparing lamotrigine to carbamazepine reported data on seizure freedom for study duration.99 The risk of remaining seizure free for the duration of study is significantly increased by 48 percent when lamotrigine is used versus carbamazepine (RR 1.48 [1.16 to 1.87]) (Appendix J Figure 27).

Two observational studies comparing levetiracetam to carbamazepine reported data on seizure freedom for study duration.103,164 The risk of remaining seizure free for the duration of study is nonsignificantly increased by seven percent when levetiracetam is used versus carbamazepine (RR 1.07 [0.93 to 1.24]) (Appendix J Figure 27).

One observational study comparing oxcarbazepine to carbamazepine reported data on seizure freedom for study duration.54 The risk of remaining seizure free for the duration of study is nonsignificantly decreased by 31 percent when oxcarbazepine is used versus carbamazepine (RR 0.69 [0.31 to 1.37]) (Appendix J Figure 27).

One observational study comparing topiramate to carbamazepine reported data on seizure freedom for study duration.105 The risk of remaining seizure free for the duration of study is nonsignificantly increased by 5 percent when topiramate is used versus carbamazepine (RR 1.05 [0.87 to 1.29]) (Appendix J Figure 27).

One observational study comparing vigabatrin to carbamazepine reported data on seizure freedom for study duration.98 The risk of remaining seizure free for the duration of study is nonsignificantly decreased by 18 percent when vigabatrin is used versus carbamazepine (RR 0.82 [0.58 to 1.13]) (Appendix J Figure 27).

Six observational studies comparing newer antiepileptic medications to carbamazepine reported data on seizure freedom for study duration.54,98,99,103,105,164 The risk of remaining seizure free for the duration of the study is nonsignificantly increased by 7 percent when newer antiepileptic medications are used versus carbamazepine (RR 1.75 [0.90 to 1.29]) (Appendix J Figure 27). A low level of statistical heterogeneity was detected (I2: 20.8 percent), but publication bias was not detected (Egger's p=0.0945).

Two randomized controlled trials reported data on seizure freedom for study duration when newer antiepileptics were compared with controlled or sustained release carbamazepine and both were amenable for pooling.81,86

One randomized controlled trial reported data on seizure freedom for study duration when lamotrigine was compared with sustained release carbamazepine.86 The risk of remaining seizure free for the study duration was nonsignificantly decreased by 18 percent when lamotrigine was used versus sustained release carbamazepine (RR 0.82 [0.64 to 1.03]) (Appendix J Figure 28).

One randomized controlled trial reported data on seizure freedom for study duration when levetiracetam was compared with controlled-release carbamazepine.81 The risk of remaining seizure free for the study duration was nonsignificantly decreased by 6 percent when levetiracetam is used versus controlled-release carbamazepine (RR 0.94 [0.80 to 1.10]) (Appendix J Figure 28).

Two randomized controlled trials reported data on seizure freedom for study duration when lamotrigine or levetiracetam were compared with controlled- or sustained-release carbamazepine and both were amenable for pooling.81,86 The risk of remaining seizure free for the study duration was nonsignificantly decreased by 10 percent when either newer agent was compared with controlled- or sustained-release carbamazepine (RR 0.90 [0.79 to 1.02]) (Appendix J Figure 28).

Phenytoin Versus Newer

Four randomized controlled trials reported seizure freedom for study duration while patients were receiving a newer antiepileptic medications compared with phenytoin57,59,64,90 and were all amenable to pooling.

One randomized controlled trial comparing lamotrigine to phenytoin reported data on seizure freedom for study duration.64 The risk of remaining seizure free for the study duration is nonsignificantly decreased by 5 percent when lamotrigine is used versus phenytoin (RR 0.95 [0.56 to 1.63]) (Appendix J Figure 29).

Two trials comparing oxcarbazepine to phenytoin reported data on seizure freedom for study duration.57,59 The risk of remaining seizure free for the study duration is nonsignificantly decreased by 3 percent when oxcarbazepine is used versus phenytoin (RR 0.97 [0.84 to 1.12]) (Appendix J Figure 29).

One trial comparing topiramate to phenytoin reported data on seizure freedom for study duration.90 The risk of remaining seizure free for the study duration is significantly decreased by 11 percent when topiramate is used versus phenytoin (RR 0.89 [0.80 to 0.98]) (Appendix J Figure 29). Given the RD (RD 0.89 [0.82 to 0.98]), for every two patients treated with topiramate, one less patient would remain seizure free for the duration of the study when treated with phenytoin.

Four trials comparing newer antiepileptic medications to phenytoin reported data on seizure freedom for study duration.57,59,64,90 The risk of remaining seizure free for the study duration is significantly decreased by 8 percent when newer antiepileptic medications are used versus phenytoin (RR 0.92 [0.85 to 1.00]) (Appendix J Figure 29). No statistical heterogeneity was detected (I2: 0 percent) or publication bias was detected (Egger's p=0.257).

One observational study comparing oxcarbazepine to phenytoin reported data on seizure freedom on study duration.54 The risk of remaining seizure free for the study duration is nonsignificantly decreased by 39 percent when newer antiepileptic medications are used versus phenytoin (RR 0.61 [0.28 to 1.15]).

Valproic Acid Versus Newer

Fifteen studies (12 randomized controlled trials and 3 observational studies) reported seizure freedom for study duration while patients were receiving newer antiepileptic medications compared with valproic acid.54,58,68,72,74,76,77,80,82,84,87,88,94,99,105

Twelve randomized controlled trials reported data on seizure freedom for study duration when newer antiepileptics were compared versus valproic acid58,68,72,74,76,77,80,82,84,87,88,94 and were all amenable to pooling.

Seven trials comparing lamotrigine to valproic acid reported data on seizure freedom for study duration.23,68,72,80,87,88,94 The risk of remaining seizure free for the study duration is nonsignificantly decreased by 5 percent when lamotrigine is used versus valproic acid (RR 0.95 [0.76 to 1.19]) (Appendix J Figure 30). A moderate level of statistical heterogeneity was detected (I2: 47.3 percent), but publication bias was not detected (Egger's p=0.132).

Two trials comparing oxcarbazepine to valproic acid reported data on seizure freedom for study duration.58,82 The risk of remaining seizure free for the study duration is nonsignificantly increased by 1 percent when oxcarbazepine is used versus valproic acid (RR 1.01 [0.81 to 1.26]) (Appendix J Figure 30).

Three trials comparing topiramate to valproic acid reported data on seizure freedom for study duration.74,77,84 The risk of remaining seizure free for the study duration is nonsignificantly increased by 8 percent when topiramate is used versus valproic acid (RR 1.08 [0.86 to 1.36]) (Appendix J Figure 30). No significant statistical heterogeneity (I2: 0 percent) or publication bias was detected (Egger's p=0.997).

Twelve trials comparing newer antiepileptic medications to valproic acid reported data on seizure freedom for study duration.58,68,72,74,76,77,80,82,84,87,88,94 The risk of remaining seizure free for the study duration was nonsignificantly decreased by 3 percent when newer antiepileptic medications are used versus valproic acid (RR 0.97 [0.87 to 1.08]) (Appendix J Figure 30). A low level of statistical heterogeneity was detected (I2: 5 percent), but statistically significant publication bias was detected (Egger's p=0.039).

Three observational studies reported seizure freedom for study duration while patients were receiving newer antiepileptic medications compared with valproic acid and all were amenable for pooling.54,99,105

One observational study comparing lamotrigine to valproic acid reported data on seizure freedom for the study duration.99 The risk of remaining seizure free for study duration is nonsignificantly increased by 5 percent when lamotrigine is used versus valproic acid (RR 1.05 [0.82 to 1.34]) (Appendix J Figure 31).

One observational study comparing oxcarbazepine to valproic acid reported data on seizure freedom for study duration.54 The risk of remaining seizure free for the study duration is nonsignificantly decreased by 21 percent when oxcarbazepine is used versus valproic acid (RR 0.79 [0.34 to 1.67]) (Appendix J Figure 31).

One observational study comparing topiramate to valproic acid reported data on seizure freedom for study duration.105 The risk of remaining seizure free for the study duration is nonsignificantly decreased by 20 percent when topiramate is used versus valproic acid (RR 0.80 [0.68 to 0.92]) (Appendix J Figure 31).

Three observational study comparing newer antiepileptic medications to valproic acid reported data on seizure freedom for study duration.54,99,105 The risk of remaining seizure free for the study duration is nonsignificantly decreased by 11 percent when newer antiepileptic medications are used versus valproic acid (RR 0.89 [0.72 to 1.10]) (Appendix J Figure 31). A moderate level of statistical heterogeneity was detected (I2: 46.3 percent), but tests for publication bias could not be performed.

Innovator Versus Generic Drug Evaluation

There were no controlled clinical trials or controlled observational studies that reported data on seizure freedom for study duration.

Secondary Seizure Injury
Innovator Versus Generic Drug Evaluation

There were no controlled clinical trials or controlled observational studies that reported secondary seizure injury.

Status Epilepticus
Innovator Versus Generic Drug Evaluation

There were no controlled clinical trials or controlled observational studies that reported secondary status epilepticus.

Discussion

The newer antiepileptic medications did not impact the risk of mortality versus their older counterparts carbamazepine, phenytoin, or valproic acid. However, many of these trials had followup times that might preclude observing an impact on a long-term outcome such as survival.

No significant difference in the risk of maintaining seizure freedom was seen when newer antiepileptic medications were compared versus carbamazepine, controlled/sustained release carbamazepine, phenytoin, or valproic acid in controlled clinical trials. However, there was a trend towards a reduction in the risk of maintaining seizure freedom when newer antiepileptic medications were compared with phenytoin and there was a significant reduction in the risk of maintaining seizure freedom when topiramate was compared with phenytoin. Data is limited for the comparisons of newer antiepileptic medications versus controlled/sustained release carbamazepine.

The risk of being seizure free for either 12 or 24 months was significantly lower for newer antiepileptic agents versus carbamazepine. In individual newer antiepileptic medication versus carbamazepine analyses, the risk of being seizure free was significantly reduced by gabapentin or oxcarbazepine versus carbamazepine at 12 and 24 months and for topiramate versus carbamazepine at 12 months. No differences in 12- or 24-month seizure freedom were seen for newer antiepileptic medications versus valproic acid, although this was based on a single controlled clinical trial.

There was a significant increase in the time to first seizure when newer antiepileptic medications were compared versus phenytoin. No difference in the time to first seizure was seen between newer antiepileptic medications versus carbamazepine or valproic acid. However, in individual newer agent versus carbamazepine or valproic acid analyses, significant reductions were seen for gabapentin and vigabatrin versus carbamazepine and for lamotrigine versus valproic acid.

Statistical heterogeneity in some of the analyses of older versus newer antiepileptic medications was likely due to the pooling of different newer antiepileptic medications together and pooling patients with generalized epilepsy, partial epilepsy, new onset epilepsy, and chronic epilepsy together. We performed a priori–defined subgroup analyses to explore heterogeneity due to these factors as well as gender and age.

For the comparison of innovator antiepileptic medications to their respective generic versions, we found that seizure occurrence and frequency was similar between groups in controlled clinical trials. In addition, there were no differences between innovator antiepileptic medications and their respective generic versions in terms of total withdrawals or withdrawals due to lack of efficacy in controlled clinical trials. In one controlled observational trial, there was a significant increase in withdrawals for any reason, but this study had marked differences in several demographic variables (age, insurance type, and concomitant migraine headache and cerebral palsy), but the study investigators did not conduct adjusted analyses.

In 2010, a meta-analysis on seizure occurrence following the use of generic versus innovator antiepileptic medications was published.165 In this meta-analysis, the authors pooled seven trials evaluating the occurrence of seizures together. We did not include the trial by Wolf 1992 since they were comparing two established versions of a sustained release carbamazepine product versus a new version. The new version was not a generic of the original versions and was not included. The authors said they included data from Hartley 1991 but instead used the data from Hartley 1990. Our findings, using the six trials that were eligible for pooling within our analysis (RR 0.89 [0.57 to 1.39]), are characteristically similar to that of the meta-analysis by Kesselheim 2010 (OR 1.1 [0.9 to 1.2]).

When viewed together, the data suggest that generic antiepileptic medication use, predominantly with carbamazepine, phenytoin, and valproic acid, provides a similar level of efficacy to a population of people with epilepsy as their respective innovator products. This occurred even though many of the trials did not use FDA approved “A” rated generics, which would have likely resulted in less variability in concentrations between the different forms of the medications. It would be difficult to extrapolate these findings from controlled clinical trials to other antiepileptic medications since they were not represented in the analyses.

Many of the controlled clinical trials used a crossover design or randomized patients to either an innovator or generic product in a parallel fashion so they cannot be used to determine whether a switch from one antiepileptic medication to another “A” rated form of the medication, whether an innovator or generic, would increase the risk of seizure occurrence or increase seizure frequency. Unfortunately, this has not been directly assessed in any controlled clinical trial or controlled observational trial. It has been reported in descriptive trials, but these uncontrolled observations are prone to such a high degree of bias, they cannot be used to reliably gauge comparative efficacy.

In the absence of controlled clinical trial data, controlled observational studies can be used to provide insight into the impact of innovator to generic switching on other endpoints. However, the inherent limitations of observational data need to be appreciated and negatively impact internal validity, even when sophisticated statistical methods are used to create more comparable control groups. Even under the best of circumstances, investigators can only control for those factors that are known or suspected to impact results, and the sample size limits the number of variables that can be controlled for. Additionally, as with the controlled clinical trials, it is hard to extrapolate the results on these studies to that of other antiepileptic medications that were not evaluated.

Four controlled observational trials have evaluated the impact of switching from one version of an antiepileptic medication (either an innovator or generic) to another version of the medication on outpatient resource utilization, hospitalization, and hospital stay duration. These controlled observational studies compared periods of innovator medication use versus periods of generic use. All of the observational studies were sponsored by the pharmaceutical industry. Two controlled observational trials (one evaluating several antiepileptic medications together as one group and another focusing on lamotrigine) found an increased incidence of utilizing outpatient resources33,129 but two other trials focusing on topiramate did not.133,162 One of the four trials found significant increases in hospitalization rates during periods of generic use, one found significant increases in both evaluated subpopulations, one trial found a trend towards an increase in the hospitalization rate, and another found a significant increase in the hospitalization rate in one subpopulation and a trend in the other. All four controlled observational studies found a significant increase in hospital length of stay.

These controlled observational studies, while well conducted, have four important limitations. First, they set their observational periods to coincide with the generic introduction of an antiepileptic medication. As such, they were evaluating patients who were likely stabilized on the innovator therapy, were switched to the generic medication and if they had an issue, were switched back. As such, these studies cannot be used to differentiate the comparative efficacy of innovator versus generic antiepileptic medications since the circumstances for which they were used is different. The studies do provide insight into the impact of switching from one medication to another version of the same medication. Second, it was not specified that the controlled observational studies were limited to “A” rated generics. If generic versions that would not meet the FDA guidance for an “A” rated generic were included, the differences between the innovator and generic groups may be greater than when limited to “A” rated versions. Third, the switch was not blinded. As such, patients and clinicians may have been aware the switch had occurred and emotional or anxiety related triggers for medical service utilization not related to the comparability of the innovator and generic products could have occurred. Fourth, the studies used claims data increasing the risk that data was missing or misclassified.

Three well-conducted controlled observational studies assessed a composite endpoint of medical service utilization. Two of the studies were supported by the pharmaceutical industry, used similar methods, had a similar composite endpoint (emergency department visit, ambulance service utilization, or hospitalization) and derived similar results. They matched for several important factors, limited the analyses to “A” rated products, and conducted subgroup analyses for some other factors found to be disparate between groups with similar results to the base case analysis. As such, these observational trials were well conducted. However, they could not control for comorbidities or changes in other medications and their associated dosages which are known to impact seizure occurrence. Only one of the studies matched for gender and the other study had a different regional distribution of patients between cases and controls. Given the enrolled population, it is difficult to assure that the case population had the same baseline risk of an acute event requiring emergency services aside from their switch between versions of the same antiepileptic medication. It would have increased the internal validity if they found a similar number of office visits in the 6 to 12 months preceding the switch between “A” rated versions of the antiepileptic medications. If the cases had more office visits preceding the switch it would suggest that the patients were not comparable. A third important case control study was conducted by Devine and colleagues and was sponsored by Express Scripts.128 In this study, significant increases in hospitalization of emergency room visits were seen in unadjusted analyses, but no significant difference was found after adjusting for confounders. Unlike the other two trials,126,127 this study's authors controlled for person's risk of epilepsy exacerbation, change in disease severity, drug interactions, poor adherence, and change in patient diagnosis.128 This suggests that the difference in magnitude between these three studies may be due to inadequate confounder adjustment. However, even with adjusting for confounders, the study by Devine and colleagues still had the same direction of effect as those of Zachry and colleagues and Rascati and colleagues.126-128 Since the controlled observational studies by Zachry and colleagues and Rascati and colleagues used a composite endpoint that included ambulance service utilization and Devine and colleagues did not, this may also explain differences in magnitude between the three studies. All three of these controlled observational trials were unblinded and used claims data increasing the risk that data was missing or misclassified.

Key Question 2. In patients with epilepsy, what is the comparative effectiveness/efficacy of antiepileptic medications on intermediate outcomes: pharmacokinetics, the comparative dose of medication needed to control seizures, and switchback rates?

Key Points

  • This Key Question focused on the innovator versus generic comparison. Pharmacokinetic data was derived from carbamazepine and to a lesser extent phenytoin, and lamotrigine studies, as there is limited ability to extrapolate to all antiepileptic medications.
  • The average maximum concentration (Cmax), minimum concentration (Cmin), time to maximum concentration (Tmax), and area under the curve (AUC) values from a population of patients receiving innovator antiepileptic medications are similar to that of their generic versions.
  • The average Steady-state concentration (Css) values from a population of patients receiving innovator antiepileptic medication are similar overall to that of their generic counterparts.
    • There was a significantly increase in the weighted mean difference for Css with generic versus innovator phenytoin.
  • While the average pharmacokinetic parameters of patients receiving innovator and their generic versions are similar, we cannot demonstrate whether or not an individual patient stabilized on an innovator or generic product will experience a marked change when switched to alternate therapy, leading to loss of efficacy or adverse events.
  • There are high rates of switching from innovator to generic versions of antiepileptic medications as evidenced by high rates of switching back from generic to innovator products.

Detailed Analysis

Study Design and Population Characteristics

Studies used to answer this Key Question will be the same as those comparing innovator to generic antiepileptic medications used to answer Key Question 1.

Outcome Evaluations

Pharmacokinetics
Maximum Concentration

Eight trials (seven randomized controlled trial and one prospective nonrandomized trial) reported the maximal blood concentrations (Cmax) of patients receiving innovator and their associated generic antiepileptic medications 110-116,122

Six randomized controlled trials reported data on Cmax in patients receiving innovator carbamazepine versus generic versions and were all suitable for meta-analysis.110-115 Only one of the trials utilized a discernable FDA “A” rated generic carbamazepine product.112 The weighted mean difference for Cmax in the generic carbamazepine group was nonsignificantly higher than the innovator carbamazepine group (WMD 0.28 [-0.17 to 0.72] mcg/mL) (Appendix J Figure 32). No statistical heterogeneity (I2: 0 percent) or publication bias (Egger's: p=0.277) was detected.

One prospective nonrandomized trial reported data on Cmax in patients receiving innovator lamotrigine against a generic version but did not use a discernable FDA “A” rated generic.116 The weighted mean difference in the generic lamotrigine group was nonsignificantly lower than the innovator lamotrigine group (WMD -0.11 [-8.82 to 8.60] mcg/mL) (Appendix J Figure 32). Statistical heterogeneity and publication bias could not be determined.

One randomized controlled trial reported data on Cmax in patients receiving innovator phenytoin versus a several generic versions but did not use a discernable FDA “A” rated generic.122 The weighted mean difference in the generic phenytoin group was nonsignificantly lower than the innovator phenytoin group (WMD -1.08 [-4.35 to 2.19] mcg/mL) (Appendix J Figure 32). No statistical heterogeneity was found (I2: 0 percent), but publication bias was detected (Egger's: p=0.001).

Eight trials, seven of which were randomized, reported data on Cmax in patients receiving innovator versus generic versions of antiepileptic medication and were all suitable for meta-analysis.110-116,122 The standardized mean difference in the generic antiepileptic medication group was nonsignificantly higher than the innovator antiepileptic medication group (SMD 0.10 [-0.13 to 0.32]) (Appendix J Figure 32). No statistical heterogeneity (I2: 0 percent) was found, but there was a trend towards detectable publication bias (Egger's: p=0.086).

The BCS Class I analysis is the same as that reported for lamotrigine (WMD -0.11 [-8.82 to 8.60] mcg/mL). Seven randomized controlled trials reported data on BCS Class II antiepileptic medications suitable for meta-analysis.110-115,122 The standardized mean difference in the generic BCS Class II antiepileptic medication group was nonsignificantly lower than the innovator BCS Class II antiepileptic medication group (SMD 0.10 [-0.13 to 0.33]) (Appendix J Figure 32). No statistical heterogeneity (I2: 0 percent) or publication bias (Egger's: p=0.103) was detected.

One randomized controlled trial reported on the comparison of an innovator to a discernable FDA “A” rated generic product, in this case carbamazepine.112 The weighted mean difference in the generic carbamazepine group was nonsignificantly higher than the innovator carbamazepine group (WMD 0.20 [-0.73 to 1.13] mcg/mL) (Appendix J Figure 32). Tests for statistical heterogeneity and publication bias could not be calculated.

Minimum Concentration

Six trials (four randomized controlled trials, one prospective before and after nonblinded trial, and one prospective nonrandomized trial) reported the minimal blood concentrations (Cmin) of patients receiving innovator and their associated generic antiepileptic medications.107,110,113-116

Five trials (four randomized controlled trials and one prospective before and after nonblinded trial) reported data on Cmin in patients receiving innovator carbamazepine versus generic versions and were all suitable for meta-analysis.107,110,113-115 None of the trials used a discernable FDA “A” rated generic carbamazepine product. The weighted mean difference in the generic carbamazepine group was nonsignificantly higher than the innovator carbamazepine group (WMD 0.15 [-0.25 to 0.56] mcg/mL) (Appendix J Figure 33). No statistical heterogeneity was found (I2: 0 percent), but publication bias was trending towards significance (Egger's: p=0.056).

One prospective nonrandomized trial reported data on Cmin in patients receiving innovator lamotrigine versus a generic version but did not use a discernable FDA “A” rated generic.116 The weighted mean difference in the generic lamotrigine group was nonsignificantly higher than the innovator lamotrigine group (WMD 0.89 [-6.07 to 7.85] mcg/mL) (Appendix J Figure 33). Statistical heterogeneity and publication bias could not be determined.

Six trials (four randomized controlled trials110,113-115 one prospective before and after nonblinded trial,107 and one prospective nonrandomized trial116) reported data on Cmax in patients receiving innovator versus generic version of antiepileptic medications and were all suitable for meta-analysis. The standardized mean difference in the generic group was nonsignificantly higher than the generic antiepileptic medication group (SMD 0.05 [-0.21 to 0.31]) (Appendix J Figure 33). No statistical heterogeneity (I2: 0 percent) or publication bias (Egger's: p=0.341) was detected.

The BCS Class I analysis is the same as that reported for lamotrigine (WMD 0.89 [-6.07 to 7.85] mcg/mL). The BCS Class II analysis is the same as reported for carbamazepine (WMD 0.15 [-0.25 to 0.56] mcg/mL) (Appendix J Figure 33). There were no discernable “A” rated generics available to analyze.

Steady-State Concentration

Seven randomized controlled trials reported steady state concentration (Css) of patients receiving innovator and their associated generic antiepileptic medications.108,109,111,113,118,120,121

Four randomized controlled trials reported data on Css in patients receiving innovator carbamazepine versus generic versions and were all suitable for meta-analysis.108,109,111,113 None of the trials utilized a discernable FDA “A” rated generic carbamazepine product. The weighted mean difference for Css in generic carbamazepine group was nonsignificantly higher than the innovator carbamazepine group (WMD 0.34 [-0.31 to 0.99]) (Appendix J Figure 34). No statistical heterogeneity (I2: 0 percent) or publication bias (Egger's: p=0.085) was detected.

Three randomized controlled trials reported data on Css in patients receiving innovator phenytoin versus one or more generic versions of phenytoin.118,120,121 None of the trials utilized a discernable FDA “A” rated generic. The weighted mean difference in the generic phenytoin group was significantly higher than the innovator phenytoin group (WMD 2.96 [0.65 to 5.28] mcg/mL) (Appendix J Figure 34). No statistical heterogeneity was found (I2: 0 percent), but publication bias was detected (Egger's: p=0.045).

Seven randomized controlled trials reported data on innovator versus generic versions of antiepileptic medications suitable for meta-analysis.108,109,111,113,118,120,121 The standardized mean difference in the generic group was nonsignificantly higher than the innovator antiepileptic medication group (SMD 0.18 [-0.09 to 0.45]) (Appendix J Figure 34). No statistical heterogeneity (I2: 0 percent) or publication bias (Egger's: p=0.300) was detected.

None of the available drugs were in BCS Class I, so analysis was not possible. The BCS Class II analysis is the same as antiepileptic medication analysis (SMD 0.18 [-0.09 to 0.45]) (Appendix J Figure 34). There were no discernable “A” rated generics available to analyze.

Time to Maximum Concentration

Five randomized controlled trials reported time to maximum concentration (Tmax) of patients receiving innovator and their associated generic antiepileptic medications.111-115

Five randomized controlled trials reported data on Tmax in patients receiving innovator carbamazepine versus generic versions and were all suitable for meta-analysis.111-115 Only one of the trials utilized a discernable FDA “A” rated generic carbamazepine product.112 The weighted mean difference for Tmax in generic carbamazepine group was the same as the innovator carbamazepine group (WMD 0.00 [-0.43 to 0.43] hours) (Appendix J Figure 35). High and significant statistical heterogeneity (I2: 60.2 percent) was detected, but publication bias (Egger's: p=0.400) was not detected. The heterogeneity was driven by the Aldenkamp 1996 trial. The weighted mean difference after exclusion of Aldenkamp 1996 was significantly lower in the innovator antiepileptic medication group (WMD -0.37 [-0.69 to -0.04]) with no statistical heterogeneity (I2=0 percent)( Appendix J Figure 35). We assessed potential reasons why the Aldenkamp 1996 trial would find a different direction of effect from the other trials. Aldenkamp had a short study duration (3 days per phase) versus the Hartley 1991 (6 weeks per phase) and Oles 1992 (3 months per phase), had an older mean age of 45 years versus the other trials (35 and 11 years), and was conducted more recently (1998 versus 1992 and 1991), respectively. Gender and country of study conduction were not likely explanations for the heterogeneity. Patients' seizure history was not reported adequately enough to allow assessment of this variable.

The “any antiepileptic” medication analysis and the BCS Class II analysis is the same as the carbamazepine analysis (WMD 0.00 [-0.43 to 0.43]) (Appendix J Figure 35). None of the available drugs were in BCS Class I, so analysis was not possible.

One randomized controlled trial reported on the comparison of an innovator to a discernable FDA “A” rated generic product, in this case carbamazepine.112 The WMD in the generic carbamazepine group was nonsignificantly lower than the innovator carbamazepine group (WMD -0.25 [-0.85 to 0.35]) (Appendix J Figure 35). Tests for statistical heterogeneity and publication bias could not be calculated.

Area Under the Curve

Eight trials (seven randomized controlled trials and one prospective nonrandomized trial) reported AUC of patients receiving innovator and their associated generic antiepileptic medications.110-116,122

Six randomized controlled trials reported data on AUC in patients receiving innovator carbamazepine to generic versions and were all suitable for meta-analysis.108,111-115 Only one of the trials utilized a discernable FDA “A” rated generic carbamazepine product.112 The WMD in the generic carbamazepine group was nonsignificantly higher than the innovator carbamazepine group (WMD 2.34 [-1.59 to 6.28]) (Appendix J Figure 36). No statistical heterogeneity (I2: 0 percent) or publication bias (Egger's: p=0.932) was detected.

One prospective nonrandomized trial reported data on AUC in patients receiving innovator lamotrigine versus a generic version but did not use a discernable FDA “A” rated generic.116 The WMD in the generic lamotrigine group was nonsignificantly lower than the innovator lamotrigine group (WMD -10.50 [-86.08 to 65.08] mcg/mL) (Appendix J Figure 36). Tests for statistical heterogeneity and publication bias could not be calculated.

One randomized controlled trial reported data on AUC in patients receiving innovator phenytoin versus a several generic versions but did not use a discernable FDA “A” rated generic.122 The WMD in the generic phenytoin group was nonsignificantly lower than the innovator phenytoin group (WMD -18.78 [-52.60 to 15.06]) (Appendix J Figure 36). No significant statistical heterogeneity (I2: 0 percent) was detected, but publication bias was detected (Egger's: p=0.048).

Eight trials (seven randomized controlled trials108,111-115,122 and one prospective nonrandomized trial116) reported data on innovator versus generic versions of antiepileptic medication suitable for meta-analysis. The SMD in the generic group was nonsignificantly higher than the innovator antiepileptic medication group (SMD 0.05 [-0.18 to 0.28]) (Appendix J Figure 36). No statistical heterogeneity (I2: 0 percent) was detected, but publication bias was detected (Egger's: p=0.004).

The BCS Class I analysis is the same as that reported for lamotrigine (WMD -10.50 [-86.08 to 65.08] mcg/mL). Seven randomized controlled trials reported data on BCS Class II antiepileptic medications suitable for meta-analysis.110-115,122 The standardized mean difference in the generic BCS Class II antiepileptic medication group was nonsignificantly higher than the innovator BCS Class II antiepileptic medication group (SMD 0.06 [-0.17 to 0.30]) (Appendix J Figure 36). No statistical heterogeneity (I2: 0 percent) was detected, but publication bias (Egger's: p=0.006) was detected.

One randomized controlled trial reported on the comparison of an innovator to a discernable FDA “A” rated generic product, in this case carbamazepine.112 The WMD in the “A” rated generic carbamazepine group was nonsignificantly higher than the innovator carbamazepine group (WMD 1.44 [-3.60 to 6.49]) (Appendix J Figure 36). Tests for statistical heterogeneity and publication bias could not be calculated.

Dose Requirements for Seizure Control

There were no controlled clinical trials or observational studies that reported data on this endpoint.

Switchback Rates

Four observational studies reported switchback rates, the percentage of patients switching back to an innovator antiepileptic medication after taking a generic version. In the first study,129 switchback rate in the stable epilepsy patients was 26.5 percent, whereas in the unstable epilepsy patients, the switchback rate was 31.1 percent. In the other three studies, switchback rates were divided by antiepileptic medications. Switchback rate for carbamazepine was 20.8 percent and 12.2 percent,33,132 divalproex was 14 percent,132 valproic acid was 20.9 percent and 23.9 percent,131,132 clobazam was 20.5 percent, 44.1 percent, and 23 percent,33,131,162 gabapentin was 30.9 percent and 19.5 percent,33,132 lamotrigine was 13 percent, 27.5 percent, and 12.4 percent33,131,132, and topiramate was 12.5 percent.132

Discussion

This section is specifically focused on innovator versus generic antiepileptic medications and does not pertain to older versus newer agents. While we evaluated the impact of using innovator versus generic antiepileptic medications on several pharmacokinetic parameters, the data was derived from carbamazepine and to a lesser extent phenytoin, and lamotrigine studies. Therfore, there is limited ability to extrapolate to all antiepileptic medications.

The average Cmax, Cmin, Tmax, and AUC values from a population of patients receiving innovator antiepileptic medications are similar to that of their generic versions in the combined and individual drug analyses. The average Css values from a population of patients receiving innovator antiepileptic medication are similar overall to that of their generic counterparts. However, there was a significantly increase in the weighted mean difference for Css with generic versus innovator phenytoin.

When taken together, a population of patients should derive similar concentrations on an innovator as they would using a generic antiepileptic medication. However, our data do not allow us to determine if an individual patient or subset of patients would have an over- or underaccentuated pharmacokinetic response if they were switched from one version of the medication to the other.

While 12 to 44 percent of patients in four observational studies switched back to innovator antiepileptics after taking a generic version of the medication, the main limitation of this type of data is that the patients and clinicians were not blinded. As a result, the switchback from a generic to an innovator antiepileptic medication may or may not be due to real versus perceived differences in efficacy or adverse events

Key Question 3. In patients with epilepsy, what is the comparative impact of antiepileptic medications on serious adverse events such as neurological adverse effects, hypotension, rash, suicidal ideation, mood and cognition, bone density, and cosmetic adverse effects?

Key Points

  • While the risk of withdrawing for any reason is not different for newer antiepileptic medications versus either carbamazepine or controlled/sustained carbamazepine in controlled clinical trials, this is due to an offsetting increase in the risk of withdrawals due to lack of efficacy and a decrease in withdrawals due to adverse events.
  • No difference was found in the risk of withdrawals for any reason or withdrawals due to adverse events when newer antiepileptic medications were compared with ethosuximide, although this is based on a single trial with lamotrigine.
  • No difference was found in the risk of withdrawals for any reason, withdrawals due to lack of efficacy, or withdrawals due to adverse events when newer antiepileptic medications were compared with either phenytoin or valproic acid, although the phenytoin comparisons were based on limited trial data with only lamotrigine and oxcarbazepine as comparators.
  • The risk of dizziness was significantly lower with newer antiepileptic medications versus carbamazepine and in individual comparisons between lamotrigine or topiramate versus carbamazepine. No differences were noted between newer antiepileptic medications versus either phenytoin or valproic acid, although the phenytoin evaluation had limited data. No data were available for controlled/sustained-release carbamazepine.
  • The risk of fatigue was significantly lower with newer antiepileptic medications versus carbamazepine and in individual comparisons between gabapentin, lamotrigine, or topiramate versus carbamazepine. Similarly, the risk of fatigue was significantly lower with newer antiepileptic medications versus valproic acid and in and individual comparison between topiramate versus carbamazepine. No differences were noted between newer antiepileptic medications versus either controlled/sustained-release carbamazepine or phenytoin, although these evaluations had limited data.
  • No difference was found in the risk of headache when newer antiepileptic medications were compared versus carbamazepine, controlled/sustained release carbamazepine, phenytoin, or valproic acid, although the phenytoin and controlled/sustained-release carbamazepine evaluations were based on limited data. The risk of headache was significantly lower when oxcarbazepine was compared versus valproic acid.
  • No difference was found in the risk of nausea when newer antiepileptic medications were compared versus carbamazepine, controlled/sustained-release carbamazepine, or phenytoin, although the controlled/sustained release carbamazepine evaluations were based on limited data and a significantly lower risk of nausea was noted in individual agent analysis when topiramate was compared versus carbamazepine. The risk of nausea was significantly lower when newer antiepileptic medications were compared versus valproic acid and in individual comparisons where lamotrigine or topiramate had a significantly lower risk of nausea than valproic acid.
  • No difference in the risk of vomiting was noted for newer antiepileptic medications versus carbamazepine or valproic acid but was significantly lower when compared versus phenytoin. The risk of vomiting was based on limited data for the carbamazepine and phenytoin comparisons, and no data was available evaluating newer antiepileptic medications and controlled/sustained-release carbamazepine.
  • The risk of skin rash was significantly lower with newer antiepileptic medications versus carbamazepine and in individual comparisons between gabapentin, lamotrigine, topiramate, or vigabatrin versus carbamazepine. Similarly, the risk of skin rash was significantly lower with newer antiepileptic medications versus controlled/sustained-release carbamazepine. No differences were noted between newer antiepileptic medications versus either phenytoin or valproic acid although in individual agent analysis, the risk of skin rash was significantly reduced for topiramate versus either phenytoin or valproic acid but was significantly increased for lamotrigine versus valproic acid.
  • The risk of somnolence was significantly lower with newer antiepileptic medications versus carbamazepine and in individual comparisons between gabapentin or lamotrigine versus carbamazepine. Similarly, the risk of somnolence was significantly lower with newer antiepileptic medications versus valproic acid. No differences were noted between newer antiepileptic medications versus either controlled/sustained-release carbamazepine or phenytoin, although the controlled/sustained-release evaluation had limited data and there was a significantly lower risk of somnolence in individual agent analysis when lamotrigine was compared versus phenytoin.
  • No differences in the risk of alopecia was noted when newer antiepileptic medications and carbamazepine. Newer antiepileptic medications decreased the risk of experiencing alopecia when compared with valproic acid and when lamotrigine and topiramate were individually compared with valproic acid.
  • No difference in the risk of acne was seen when newer antiepileptic medications were compared versus phenytoin, but this was based on a single study. However, newer antiepileptic medications (only oxcarbazepine was evaluated) had a lower risk of gum hyperplasia when compared with phenytoin.
  • Cognition was evaluated for several newer antiepileptic medications versus carbamazepine and valproic acid with very limited data with phenobarbital and phenytoin.
  • Mood was not as extensively evaluated as cognition, and in several places, lack of significant effects was noted but data were not provided.
  • No clinical trial evidence was available evaluating bone mineral density in patients receiving older and newer antiepileptic medications. These studies have higher risk of inherent biases and may be underpowered.
    • Once study found that the use of either newer or older agents decreased bone mineral density versus normal controls. One study found no difference in bone mineral density when lamotrigine was compared against older antiepileptic drugs (carbamazepine, phenytoin, valproic acid). A final study found that carbamazepine reduced bone mineral density versus baseline but valproic acid and lamotrigine did not.
  • Data on withdrawal rates due to adverse events were only available for innovator versus generic carbamazepine and phenytoin, limiting the ability to extrapolate findings to other antiepileptic medications.
  • The withdrawals due to adverse events were similar between the innovator and generic versions of antiepileptic medications.

Detailed Analysis

Study Design and Population Characteristics

Older Versus Newer Antiepileptic Drug Evaluation

Studies to answer Key Question 3 are derived from the same set of studies used to evaluate Key Question 1 and are summarized in Appendix G Tables 15-19.

Innovator Versus Generic Antiepileptic Drug Evaluation

Studies to answer Key Question 3 are derived from the same set of studies used to evaluate Key Question 1 and are summarized in Appendix G Tables 5, 7-9.

Outcome Evaluations

Withdrawals Due to Adverse Events
Older Versus Newer Antiepileptic Drug Evaluation
Carbamazepine Versus Newer

Twenty-four trials (18 randomized controlled trials and 6 observational studies) reported withdrawals due to adverse events while patients were receiving a newer antiepileptic medication compared with carbamazepine.48,49,52,53,55,56,60,62,63,70,74,76-80,83,85,95,98,99,101,103,104

Eighteen randomized controlled trials reported data on withdrawals due to adverse events when newer antiepileptics were compared versus carbamazepine and were amenable for pooling. 48,49,52,53,55,56,60,62,63,70,74,76-80,83,85

Three randomized controlled trials reported data on withdrawals due to adverse events when gabapentin was compared versus carbamazepine and were amenable for pooling.60,78,85 The risk of withdrawal was significantly decreased by 49 percent when gabapentin was compared versus carbamazepine (RR 0.51 [0.33 to 0.79]) (Appendix J Figure 37). Given the RD, (RD -0.12 [-0.17 to -0.08]), for every nine patients treated with gabapentin, one fewer patient would withdraw due to an adverse event compared with those treated with carbamazepine. A higher level of statistical heterogeneity was detected (I2: 56.5 percent), but publication bias was not detected (Egger's: p=0.089).

Eight randomized controlled trials reported data on withdrawals due to adverse events when lamotrigine was compared versus carbamazepine and were amenable for pooling.52,55,62,70,76,78,80,85 The risk of withdrawal was significantly decreased by 48 percent when lamotrigine was compared versus carbamazepine (RR 0.52 [0.43 to 0.61]) (Appendix J Figure 37). Given the RD (RD -0.17 [-0.14 to -0.07]), for every six patients treated with lamotrigine, one fewer patient would withdraw due to an adverse event compared with those treated with carbamazepine. No statistical heterogeneity (I2: 0 percent) or publication bias (Egger's: p=0.222) was detected.

Three randomized controlled trials reported data on withdrawal due to adverse events when oxcarbazepine was compared versus carbamazepine and were amenable for pooling.48,49,85 The risk of withdrawal was nonsignificantly decreased by 24 percent when oxcarbazepine was compared versus carbamazepine (RR 0.76 [0.56 to 1.04]) (Appendix J Figure 37). A lower level of statistical heterogeneity was detected (I2: 9.1 percent), but tests for publication bias could not be performed.

Four randomized controlled trials reported data on withdrawals due to adverse events when topiramate was compared versus carbamazepine and were amenable for pooling.74,77,83,85 The risk of withdrawal was nonsignificantly increased by 4 percent when topiramate was used versus carbamazepine (RR 1.04 [0.86 to 1.26]) (Appendix J Figure 37). No statistical heterogeneity (I2: 0 percent), or publication bias was detected (Egger's: p=0.496).

Four randomized controlled trials reported data on withdrawals due to adverse events when vigabatrin was compared versus carbamazepine.53,56,63,79 The risk of withdrawal was nonsignificantly decreased by 42 percent when vigabatrin was used versus carbamazepine (RR 0.58 [0.23 to 1.45]) (Appendix J Figure 37). A moderate level of statistically significant statistical heterogeneity was detected (I2: 42.7 percent), but publication bias was not detected (Egger's: p= 0.442).

In the pooled analysis of 18 randomized trials reporting data on withdrawals due to adverse events when either gabapentin, lamotrigine, oxcarbazepine, topiramate, or vigabatrin were compared versus carbamazepine.48,49,52,53,55,56,60,62,63,70,74,76-80,83,85 The risk of withdrawal was significantly decreased by 37 percent when all newer antiepileptics were used versus carbamazepine (RR 0.63 [0.53 to 0.73]) (Appendix J Figure 37). Given the RD (RD -0.08 [-0.11 to -0.053]), for every 13 patients treated with a newer agent, 1 fewer patient would withdraw due to an adverse event compared with those treated with carbamazepine. A moderate level of statistically significant statistical heterogeneity was detected (I2: 35.4 percent), but publication bias was not detected (Egger's: p= 0.063).

Six observational studies reported withdrawals due to adverse events while patients were receiving newer antiepileptic medications compared with carbamazepine and all six were amenable for pooling.95,98,99,101,103,104

Two observational studies reported withdrawals due to adverse events when lamotrigine was compared versus carbamazepine and were amenable for pooling.99,101 The risk of withdrawal was significantly decreased by 60 percent when lamotrigine was compared versus carbamazepine (RR 0.40 [0.20 to 0.78]) (Appendix J Figure 38). Given the RD, (RD -0.10 [-0.27 to 0.07]), for every 10 patients treated with lamotrigine, 1 less patient would withdraw compared with those treated with carbamazepine.

One observational study reported withdrawal due to adverse events when levetiracetam was compared versus carbamazepine.103 The risk of withdrawal was nonsignificantly increased by 2.4-fold when levetiracetam was compared versus carbamazepine (RR 2.42 [0.45 to 14.69]) (Appendix J Figure 38).

One observational study reported withdrawal due to adverse events when oxcarbazepine was compared versus carbamazepine.95 The risk of withdrawal was nonsignificantly increased by 3-fold when oxcarbazepine was compared versus carbamazepine (RR 3.00 [0.26 to 35.71]) (Appendix J Figure 38).

One observational study reported withdrawal due to adverse events when topiramate was compared versus carbamazepine.104 The risk of withdrawal was nonsignificantly decreased by 63 percent when topiramate was compared versus carbamazepine (RR 0.37 [0.06 to 2.14]) (Appendix J Figure 38).

One observational study reported withdrawal due to adverse events when vigabatrin was compared versus carbamazepine.98 The risk of withdrawal was nonsignificantly increased by 8 percent when vigabatrin was compared versus carbamazepine (RR 1.08 [0.06 to 18.74]) (Appendix J Figure 38).

In the pooled analysis of six observational studies reporting data on withdrawals due to adverse events, either lamotrigine, levetiracetam, oxcarbazepine, topiramate, or vigabatrin was compared versus carbamazepine.95,98,99,101,103,104 The risk of withdrawal was significantly increased by 49 percent when either newer agent was compared versus carbamazepine (RR 0.51 [0.28 to 0.92]) (Appendix J Figure 38). Given the RD (RD -0.020 [-0.098 to 0.057]), for every 50 patients treated with a newer agent, 1 less patient would withdraw compared with those treated with carbamazepine. No significant statistical heterogeneity (I2: 0 percent) or publication (Egger's: p=0.098) was detected.

Two randomized controlled trials reported withdrawals due to adverse events when newer antiepileptics were compared with controlled- or sustained-release carbamazepine and both were amenable for pooling.81,86

One randomized controlled trial reported withdrawals due to adverse events when lamotrigine was compared with sustained-release carbamazepine.86 The risk of withdrawal was nonsignificantly decreased by 45 percent when lamotrigine was compared with sustained-release carbamazepine (RR 0.55 [0.30 to 1.01]) (Appendix J Figure 39).

One randomized controlled trial reported withdrawals due to adverse events when levetiracetam was compared versus controlled-release carbamazepine.81 The risk of withdrawal was nonsignificantly decreased by 25 percent when levetiracetamwas compared with controlled-release carbamazepine (RR 0.75 [0.52 to 1.08]).

Two randomized controlled trials reported withdrawals due to adverse events when lamotrigine or levetiracetam was compared with controlled- or sustained-release carbamazepine and both were amenable for pooling.81,86 The risk of withdrawal was significantly decreased by 31 percent when either newer agent was compared with controlled- or sustained-release carbamazepine (RR 0.69 [0.50 to 0.95]) (Appendix J Figure 39). Given the RD (RD -0.0063 [-0.012 to -0.0090]), for every 159 patients treated with a newer agent, 1 less patient would withdraw compared with those treated with controlled or sustained release carbamazepine.

Phenytoin Versus Newer

Three randomized controlled trials reported withdrawals due to adverse events while patients were receiving a newer antiepileptic medication compared with phenytoin and all three were amenable for pooling.57,59,64

One randomized controlled trial reported data on withdrawals due to adverse events when lamotrigine was compared with phenytoin.64 The risk of withdrawal was nonsignificantly decreased by 20 percent when lamotrigine was compared versus phenytoin (RR 0.80 [0.42 to 1.51]) (Appendix J Figure 40).

Two randomized controlled trials reported data on withdrawals due to adverse events when oxcarbazepine was compared with phenytoin and were amenable for pooling.57,59 The risk of withdrawal was significantly decreased by 75 percent when oxcarbazepine was compared versus phenytoin (RR 0.25 [0.11 to 0.55]) (Appendix J Figure 40). Given the RD (RD -0.095 [-0.14 to -0.05]), for every 11 patients treated with oxcarbazepine, 1 less would withdraw due to adverse events compared with those treated with carbamazepine.

In the pooled analysis of three randomized trials reporting data on withdrawals due to adverse events either lamotrigine or oxcarbazepine were compared versus phenytoin and were amenable for pooling.57,59,64 The risk of withdrawal was nonsignificantly decreased by 62 percent when the newer agents were compared versus phenytoin (RR 0.38 [0.14 to 1.03]) (Appendix J Figure 40). Significant statistical heterogeneity was detected (I2: 66.8 percent), but tests for publication bias could not be performed.

Valproic Acid Versus Newer

Eighteen studies (16 randomized controlled trials and 2 observational studies) reported withdrawals due to adverse events while patients were receiving a newer antiepileptic medication compared with valproic acid.58,62,68,72-77,80,84,85,87-89,94,99,101

Sixteen randomized controlled trials reported data on withdrawals due to adverse events when newer antiepileptics were compared versus valproic acid and all were amenable for pooling.58,61,68,72-77,80,84,85,87-89,94

Ten randomized controlled trials reported data on withdrawals due to adverse events when lamotrigine was compared versus valproic acid and were amenable for pooling.68,72,75,76,80,85,87-89,94 The risk of withdrawal was significantly decreased by 28 percent when lamotrigine was compared versus valproic acid (RR 0.72 [0.57 to 0.91]) (Appendix J Figure 41). Given the RD (RD -0.02 [-0.05 to 0.004]), for every 50 patients treated, 1 less patient would withdrawal overall on lamotrigine therapy compared versus valproic acid therapy. No statistical heterogeneity (I2: 0 percent) or publication bias (Egger's: p=0.066) was detected.

One randomized controlled trial reported data on withdrawals due to adverse events when oxcarbazepine was compared versus valproic acid.58 The risk of withdrawal was nonsignificantly increased by 42 percent oxcarbazepine when was compared versus valproic acid (RR 1.42 [0.68 to 2.99]) (Appendix J Figure 41).

Five randomized controlled trials reported data on withdrawals due to adverse events when topiramate was compared versus valproic acid and were amenable for pooling.73,74,77,84,85 The risk of withdrawal was nonsignificantly increased by 6 percent when topiramate was compared versus valproic acid (RR 1.06 [0.70 to 1.61]) (Appendix J Figure 41). A moderate level of statistical heterogeneity was detected (I2: 40.6 percent) however, publication bias was not detected (Egger's: p=0.153).

One randomized controlled trial reported data on withdrawals due to adverse events when vigabatrin was compared versus valproic acid.61 The risk of withdrawal was nonsignificantly decreased by 1 percent when vigabatrin was compared versus valproic acid (RR 0.9 [0.47 to 2.07]) (Appendix J Figure 41).

In the pooled analysis of 16 randomized controlled trials reporting data on withdrawals due to adverse events when lamotrigine, oxcarbazepine, topiramate, or vigabatrin were compared versus valproic acid.58,61,68,72-77,80,84,85,87-89,94 The risk of withdrawal was nonsignificantly decreased by 10 percent when the newer agents were compared versus valproic acid in patients (RR 0.90 [0.75 to 1.08]) (Appendix J Figure 41). A low level of statistical heterogeneity (I2: 3.1 percent) was detected, but publication bias was not detected (Egger's: p=0.840).

Ethosuximide Versus Newer

One randomized controlled trial reported withdrawals due to adverse events while patients were receiving lamotrigine compared with ethosuximide.89 The risk of overall withdrawal was nonsignificantly decreased by 29 percent when lamotrigine was compared versus ethosuximide (RR 0.71 [0.45 to 1.12]).

Innovator Versus Generic Drug Evaluation

Ten trials (nine randomized controlled trials108-113,115,120,121 and one prospective nonrandomized trial)117 reported withdrawals due to adverse events while patients were receiving innovator and their associated generic antiepileptic medications.

Seven randomized controlled trials reported on withdrawals due to adverse events for innovator carbamazepine versus one or more generic versions and were suitable for meta-analysis.108-113,115 Only one of the trials utilized a discernable Food and Drug Administration (FDA) “A” rated generic carbamazepine product.112 The risk of withdrawals due to adverse effects is nonsignificantly decreased by 15 percent when generic carbamazepine was used versus innovator carbamazepine (RR 0.85 [0.25 to 2.88]) (Appendix J Figure 43). No significant statistical heterogeneity (I2: 0 percent) or publication bias was detected (Egger's: p=0.5877).

Three trials (two randomized controlled trials,120,121 and one prospective nonrandomized trial117) reported on withdrawals due to adverse events for innovator phenytoin versus three generic versions and were all suitable for meta-analysis. None of the trials used discernable FDA “A” rated generics. The risk of withdrawals due to adverse effects is nonsignificantly decreased by 34 percent when generic phenytoin was used versus innovator phenytoin (RR 0.66 [0.10 to 4.42]) (Appendix J Figure 43). No significant statistical heterogeneity (I2: 0 percent) or publication bias was detected (Egger's: p=0.108).

Ten trials (nine randomized controlled trials,108-113,115,120,121 and one prospective nonrandomized trial117) reported on withdrawals due to adverse events for any innovator versus generic versions of antiepileptic medication and they were all suitable for meta-analysis. The risk of withdrawals due to adverse effects is nonsignificantly decreased by 21 percent when generic antiepileptic medications were used versus their appropriate innovator antiepileptic medications (RR 0.79 [0.28 to 2.20]) (Appendix J Figure 43). No significant statistical heterogeneity (I2: 0 percent) or publication bias was detected (Egger's: p=0.282).

None of the available drugs were in BCS Class I, so analysis was not possible. The BCS Class II analysis is the same as any analysis for any antiepileptic medications (RR 0.79 [0.28 to 2.20]) (Appendix J Figure 43).

One randomized controlled trial reported on the comparison of an innovator to a discernable FDA “A” rated generic product, in this case carbamazepine.112 The risk of withdrawals due to adverse effects is equivalent when “A” rated generic carbamazepine was used versus innovator carbamazepine (RR 1.00 [0.11 to 9.22]) (Appendix J Figure 43).

Neurological Adverse Events
Headache
Older Versus Newer Antiepileptic Drug Evaluation
Carbamazepine Versus Newer

Seventeen studies (15 randomized controlled trials and 2 observational studies) reported headache while patients were receiving a newer antiepileptic medications compared with carbamazepine.47,52,53,55,56,60,62,63,70,74,76-78,82,85,99,105

Fifteen randomized controlled trials reported data on headache when newer antiepileptics were compared versus carbamazepine47,52,53,55,56,60,62,63,70,74,76-78,82,85 and were all amenable to pooling.

Three trials comparing gabapentin to carbamazepine reported data on headache.60,78,85 Risk of headache is nonsignificantly decreased by 7 percent when gabapentin is used versus carbamazepine (RR 0.93 [0.67 to 1.28]) (Appendix J Figure 44). No statistical heterogeneity was detected (I2: 0 percent), but significant publication bias was detected (Egger's p=0.046).

Seven trials comparing lamotrigine to carbamazepine reported data on headache.52,55,62,70,76,78,85 Risk of headache is nonsignificantly increased by 6 percent when lamotrigine is used versus carbamazepine (RR 1.06 [0.84 to 1.33]) (Appendix J Figure 44). A low level of statistical heterogeneity (I2: 13.8 percent) was detected, but no publication bias was detected (Egger's p=0.880).

Three trials comparing oxcarbazepine to carbamazepine reported data on headache.47,82,85 Risk of headache is nonsignificantly decreased by 15 percent when oxcarbazepine is used versus carbamazepine (RR 0.85 [0.44 to 1.66]) (Appendix J Figure 44). No statistical heterogeneity was detected (I2: 0 percent) and tests for publication bias could not be performed.

Three trials comparing topiramate to carbamazepine reported data on headache.74,77,85 Risk of headache is significantly decreased by 37 percent when topiramate is used versus carbamazepine (RR 0.63 [0.41 to 0.96]) (Appendix J Figure 44). Given the RD (RD -0.057 [-0.144 to 0.030]), for every 18 patients treated, 1 less patient would develop headache from treatment with topiramate than with carbamazepine. A low level of statistical heterogeneity (I2: 33.3 percent) was detected, but no publication bias was detected (Egger's p=0.308).

Three trials comparing vigabatrin to carbamazepine reported data on headache.53,56,63 Risk of headache is nonsignificantly increased by 13 percent when vigabatrin is used versus carbamazepine (RR 1.13 [0.55 to 2.32]) (Appendix J Figure 44). A low level of statistical heterogeneity was detected (I2: 10.5 percent), but tests for publication bias could not be performed.

Fifteen trials comparing newer antiepileptic medications to carbamazepine reported data on headache.47,52,53,55,56,60,62,63,70,74,76-78,82,85 Risk of headache is nonsignificantly decreased by 8 percent when newer antiepileptic medications are used versus carbamazepine (RR 0.92 [0.78 to 1.08]) (Appendix J Figure 44). A low level of statistical heterogeneity was detected (I2: 8.6 percent), but no publication bias was detected (Egger's p=0.689).

Five observational studies reported data on headache while patients were receiving a newer antiepileptic medication compared with carbamazepine and all five were amenable for pooling.95,99,103-105

One observational study comparing lamotrigine to carbamazepine reported data on headache.99 Risk of headache is nonsignificantly decreased by 55 percent when lamotrigine is used versus carbamazepine (RR 0.45 [0.11 to 1.74]) (Appendix J Figure 45).

One observational study comparing topiramate to carbamazepine reported data on headache.105 Risk of headache is nonsignificantly decreased by 58 percent when topiramate is used versus carbamazepine (RR 0.42 [0.12 to 1.50]) (Appendix J Figure 45).

Two observational studies comparing newer antiepileptic medications to carbamazepine reported data on headache.99,105 Risk of headache is nonsignificantly decreased by 57 percent when newer antiepileptic medications are used versus carbamazepine (RR 0.43 [0.16 to 1.18]) (Appendix J Figure 45).

Two randomized controlled trials reported data on headache when newer antiepileptics were compared with controlled- or sustained-release carbamazepine and both were amenable for pooling.81,86

One randomized controlled trial reported data on headache when lamotrigine was compared with sustained-release carbamazepine.86 The risk of headache was nonsignificantly decreased by 1 percent when lamotrigine was compared versus sustained-release carbamazepine (RR 0.99 [0.44 to 2.22]) (Appendix J Figure 46).

One randomized controlled trial reported data on headache when levetiracetam was compared with controlled-release carbamazepine.81 The risk of headache was nonsignificantly decreased by 19 percent when levetiracetam was compared versus controlled-release carbamazepine (RR 0.81 [0.60 to 1.10]) (Appendix J Figure 46).

Two randomized controlled trials reported data on headache when lamotrigine or levetiracetam were compared with controlled- or sustained-release carbamazepine and both were amenable for pooling.81,86 The risk of headache was nonsignificantly decreased by 17 percent when either newer agent was compared with controlled- or sustained-release carbamazepine [RR 0.83 (0.63 to 1.10)] (Appendix J Figure 46).

Phenytoin Versus Newer. Four trials reported data on headache while patients were receiving a newer antiepileptic medications compared with phenytoin57,59,64,90 and were all amenable to pooling.

One trial comparing lamotrigine to phenytoin reported data on headache.64 Risk of headache is nonsignificantly decreased by 45 percent when lamotrigine is used versus phenytoin (RR 0.55 [0.26 to 1.14]) (Appendix J Figure 47).

Two trials comparing oxcarbazepine to phenytoin reported data on headache.57,59 Risk of headache is nonsignificantly decreased by 18 percent when oxcarbazepine is used versus phenytoin (RR 0.82 [0.54 to 1.25]) (Appendix J Figure 47).

One trial comparing topiramate to phenytoin reported data on headache.90 Risk of headache is nonsignificantly decreased by 29 percent when topiramate is used versus phenytoin (RR 0.71 [0.34 to 1.45]) (Appendix J Figure 47).

Four trials comparing newer antiepileptic medications to phenytoin reported data on headache.57,59,64,90 Risk of headache is nonsignificantly decreased by 26 percent when newer antiepileptic medications are used versus phenytoin (RR 0.74 [0.53 to 1.02]) (Appendix J Figure 47). No statistical heterogeneity (I2: 0 percent) or publication bias was detected (Egger's p=0.623).

Valproic Acid Versus Newer

Seventeen (15 randomized controlled trials and 2 observational studies) reported headache while patients were receiving a newer antiepileptic medications compared with valproic acid.50,51,58,68,72,74-77,82,84,85,88,89,94,99,105

Fifteen trials reported data on headache when newer antiepileptics were compared versus valproic acid50,51,58,68,72,74-77,82,84,85,88,89,94 and were all amenable to pooling.

One trial comparing felbamate to valproic acid reported data on headache.51 Risk of headache is nonsignificantly decreased by 66 percent when felbamate is used versus valproic acid (RR 0.34 [0.11 to 1.07]) (Appendix J Figure 48).

Eight trials comparing lamotrigine to valproic acid reported data on headache.68,72,75,76,85,88,89,94 Risk of headache is nonsignificantly increased by 3 percent when lamotrigine is used versus valproic acid (RR 1.03 [0.73 to 1.45]) (Appendix J Figure 48). A low level of statistical heterogeneity was detected (I2: 6.8 percent), but no significant publication bias was detected (Egger's p=0.158).

Two trials comparing oxcarbazepine to valproic acid reported data on headache.58,82 Risk of headache is significantly decreased by 46 percent when oxcarbazepine is used versus valproic acid (RR 0.54 [0.32 to 0.93]) (Appendix J Figure 48). Given the RD (RD -0.083 [-0.160 to -0.007]), for every 13 patients treated, 1 less patient would develop headache from treatment with oxcarbazepine than with valproic acid.

Four trials comparing topiramate to valproic acid reported data on headache.74,77,84,85 Risk of headache is nonsignificantly decreased by 6 percent when topiramate is used versus valproic acid (RR 0.94 [0.62 to 1.43]) (Appendix J Figure 48). No statistical heterogeneity was detected (I2: 0 percent), but statistically significant publication bias was detected (Egger's p=0.048).

Fifteen trials comparing newer antiepileptic medications to valproic acid reported data on headache.50,51,58,68,72,74-77,82,84,85,88,89,94 Risk of headache is nonsignificantly decreased by 10 percent when newer antiepileptic medications are used versus valproic acid (RR 0.90 [0.70 to 1.16]) (Appendix J Figure 48). A low level of statistical heterogeneity (I2: 12.9 percent) and a trend towards significant publication bias was detected (Egger's p=0.082).

Two observational studies reported on headache while patients were receiving a newer antiepileptic medication compared with valproic acid and all were amenable for pooling.99,105

One observational study comparing lamotrigine to valproic acid reported data on headache.99 Risk of headache is nonsignificantly increased by 36 percent when lamotrigine is used versus valproic acid (RR 1.36 [0.29 to 6.19]) (Appendix J Figure 49).

One observational study comparing topiramate to valproic acid reported data on headache.105 Risk of headache is nonsignificantly decreased by 45 percent when topiramate is used versus valproic acid (RR 0.55 [0.16 to 1.82]) (Appendix J Figure 49).

Two observational studies comparing newer antiepileptic medications to valproic acid reported data on headache.99,105 Risk of headache is nonsignificantly decreased by 23 percent when newer antiepileptic medications are used versus valproic acid (RR 0.77 [0.28 to 2.13]) (Appendix J Figure 49).

Ethosuximide Versus Newer

One randomized controlled trial reported data on headache while patients were receiving lamotrigine compared with ethosuximide.89 The risk of headache was nonsignificantly decreased by 34 percent when lamotrigine was compared versus ethosuximide (RR 0.66 [0.33 to 1.29]).

Innovator Versus Generic Drug Evaluation

Two randomized controlled trials reported data on headache when generic carbamazepine was compared to innovator carmabazepine.109,110 Headache is nonsignificantly increased by 4 percent when generic carbamazepine was used versus innovator carbamazepine (RR 1.04 [0.59 to 1.82]). Headache was nonsignificantly decreased by 68 percent when generic phenytoin was used versus innovator phenytoin (RR 0.32 [0.04 to 2.40]). Analysis for any antiepileptic drug showed that headache was nonsignificantly decreased by 5 percent when generic antiepileptic medications were used versus innovator antiepileptic medications (RR 0.95 [0.55 to 1.64]). BCS Class II antiepileptic medications analysis is the same as any antiepileptic drug analysis. No data was available to perform BCS Class I antiepileptic medications analysis.

Fatigue
Older Versus Newer Antiepileptic Drug Evaluation
Carbamazepine Versus Newer

Seven randomized controlled trials reported data on fatigue when newer antiepileptics were compared versus carbamazepine56,60,63,74,77,80,82 and were all amenable to pooling.

One trial comparing gabapentin to carbamazepine reported data on fatigue.60 Risk of fatigue is significantly decreased by 63 percent when gabapentin is used versus carbamazepine (RR 0.37 [0.22 to 0.62]) (Appendix J Figure 50). Given the RD (RD -0.188 [-0.300 to -0.076]), for every six patients treated, one less patient would develop fatigue from treatment with gabapentin than with carbamazepine.

One trial comparing lamotrigine to carbamazepine reported data on fatigue.80 Risk of fatigue is significantly decreased by 66 percent when lamotrigine is used versus carbamazepine (RR 0.34 [0.20 to 0.58]) (Appendix J Figure 50). Given the RD (RD -0.284 [-0.411 to -0.157]), for every four patients treated, one less patient would develop fatigue from treatment with lamotrigine than with carbamazepine.

One trial comparing oxcarbazepine to carbamazepine reported data on fatigue.82 Risk of fatigue is nonsignificantly decreased by 11 percent when oxcarbazepine is used versus carbamazepine (RR 0.89 [0.31 to 2.70]) (Appendix J Figure 50).

Two trials comparing topiramate to carbamazepine reported data on fatigue.74,77 Risk of fatigue is significantly decreased by 46 percent when topiramate is used versus carbamazepine (RR 0.54 [0.35 to 0.83]) (Appendix J Figure 50). Given the RD (RD -0.099 [-0.172 to -0.027]), for every 11 patients treated, 1 less patient would develop fatigue from treatment with topiramate than with carbamazepine. A low level of statistical heterogeneity was detected (I2: 10.5 percent), but no significant publication bias was detected (Egger's p=0.092).

Two trials comparing vigabatrin to carbamazepine reported data on fatigue.56,63 Risk of fatigue is nonsignificantly increased by 2 percent when vigabatrin is used versus carbamazepine (RR 1.02 [0.59 to 1.76]) (Appendix J Figure 50).

Seven trials comparing newer antiepileptic medications to carbamazepine reported data on fatigue.56,60,63,74,77,80,82 Risk of fatigue is significantly decreased by 43 percent when newer AEDs are used versus carbamazepine (RR 0.57 [0.41 to 0.80]) (Appendix J Figure 50). Given the RD (RD -0.098 [-0.167 to -0.029]), for every 11 patients treated, 1 less patient would develop fatigue from treatment with newer AED than with carbamazepine. A high level of statistical heterogeneity was detected (I2: 50.2 percent), but no publication bias was seen (Egger's p=0.952).

One randomized controlled trial reported data on fatigue while patients were receiving levetiracetam compared with controlled-release carbamazepine.81 Risk of fatigue is nonsignificantly increased by 17 percent when levetiracetam is used versus controlled-release carbamazepine (RR 1.17 [0.80 to 1.72]).

Phenytoin Versus Newer

One randomized controlled trials reported fatigue while patients were receiving a newer antiepileptic medication (topiramate) compared with phenytoin.90 Risk of fatigue is nonsignificantly increased by 5 percent when topiramate is used versus phenytoin (RR 1.05 [0.49 to 2.25]).

Valproic Acid Versus Newer. Eight randomized controlled trials reported data on fatigue when newer antiepileptics were compared versus valproic acid50,51,74,77,80,82,84,89 and were all amenable to pooling.

Two trials comparing felbamate to valproic acid reported data on fatigue.50,51 Risk of fatigue is nonsignificantly decreased by 18 percent when felbamate is used versus valproic acid (RR 0.82 [0.13 to 5.10]) (Appendix J Figure 51).

Two trials comparing lamotrigine to valproic acid reported data on fatigue.80,89 Risk of fatigue is nonsignificantly decreased by 32 percent when lamotrigine is used versus valproic acid (RR 0.68 [0.37 to 1.23]) (Appendix J Figure 51).

One trial comparing oxcarbazepine to valproic acid reported data on fatigue.82 Risk of fatigue is nonsignificantly increased by 85 percent when oxcarbazepineis used versus valproic acid (RR 1.85 [0.48 to 7.58]) (Appendix J Figure 51).

Three trials comparing topiramate to valproic acid reported data on fatigue.74,77,84 Risk of fatigue is significantly decreased by 49 percent when topiramate is used versus valproic acid (RR 0.51 [0.33 to 0.78]) (Appendix J Figure 51). Given the RD (RD -0.111 [-0.196 to -0.025]), for every 10 patients treated, 1 less patient would develop fatigue from treatment with topiramate than with valproic acid. No statistical heterogeneity (I2: 0 percent) or publication bias was detected (Egger's p=0.475).

Eight trials comparing newer antiepileptic medications to valproic acid reported data on fatigue.50,51,74,77,80,82,84,89 Risk of fatigue is significantly decreased by 39 percent when newer antiepileptic medications are used versus valproic acid (RR 0.61 [0.44 to 0.85]) (Appendix J Figure 51). Given the RD (RD -0.045 [-0.085 to -0.004]), for every 23 patients treated, 1 less patient would develop fatigue from treatment with newer AED than with valproic acid. No statistical heterogeneity (I2: 0 percent) or publication bias was detected (Egger's p=0.459).

Ethosuximide Versus Newer

One randomized controlled trial reported data on fatigue while patients were receiving lamotrigine compared with ethosuximide.89 The risk of fatigue was nonsignificantly decreased by 10 percent when lamotrigine was compared versus ethosuximide (RR 0.90 [0.45 to 1.80]).

Innovator Versus Generic Drug Evaluation

Data given in general adverse event section below.

Somnolence
Older Versus Newer Antiepileptic Drug Evaluation
Carbamazepine Versus Newer

Twelve studies (eight randomized controlled trials and four observational studies) reported somnolence while patients were receiving a newer antiepileptic medications compared with carbamazepine.52,55,60,62,70,76,77,79,83,99,103,105

Eight trials reported data on somnolence when newer antiepileptics were compared versus carbamazepine52,55,60,62,70,76,77,83 and were all amenable to pooling.

One trial comparing gabapentin to carbamazepine reported data on somnolence.60 Risk of somnolence is significantly decreased by 59 percent when gabapentin is used versus carbamazepine (RR 0.41 [0.18 to 0.93]) (Appendix J Figure 52). Given the RD (RD -0.081 [-0.165 to 0.002]), for every 13 patients treated, 1 less patient would develop somnolence from treatment with gabapentin than with carbamazepine. No statistical heterogeneity was detected (I2: 0 percent), but tests for publication bias could not be performed.

Five trials comparing lamotrigine to carbamazepine reported data on somnolence.52,55,62,70,76 Risk of somnolence is significantly decreased by 56 percent when lamotrigine is used versus carbamazepine (RR 0.44 [0.33 to 0.58]) (Appendix J Figure 52). Given the RD (RD -0.085 [-0.117 to -0.052]), for every 12 patients treated, 1 less patient would develop somnolence from treatment with lamotrigine than with carbamazepine. No statistical heterogeneity (I2: 0 percent) or publication bias was detected (Egger's p=0.163).

Two trials comparing topiramate to carbamazepine reported data on somnolence.77,83 Risk of somnolence is the same when topiramate is used versus carbamazepine (RR 1.00 [0.43 to 2.30]) (Appendix J Figure 52). No significant statistical heterogeneity was detected (I2: 0 percent), but tests for publication bias could not be performed.

Eight trials comparing newer antiepileptic medications to carbamazepine reported data on somnolence.52,55,60,62,70,76,77,83 Risk of somnolence is significantly decreased by 53 percent when newer antiepileptic medications are used versus carbamazepine (RR 0.47 [0.36 to 0.61]) (Appendix J Figure 52). Given the RD (RD -0.075 [-0.104 to -0.046]), for every 14 patients treated, 1 less patient would develop somnolence from treatment with newer AED than with carbamazepine. No statistical heterogeneity (I2: 0 percent) or publication bias was detected (Egger's p=0.784).

Four observational studies reported on somnolence while patients were receiving a newer antiepileptic medication compared with carbamazepine and all were amenable for pooling.79,99,103,105

One observational study comparing lamotrigine to carbamazepine reported data on somnolence.99 Risk of somnolence is nonsignificantly decreased by 59 percent when lamotrigine is used versus carbamazepine (RR 0.41 [0.04 to 3.97]) (Appendix J Figure 53).

One observational study comparing levetiracetam to carbamazepine reported data on somnolence.103 Risk of somnolence is significantly decreased by 77 percent when levetiracetam is used versus carbamazepine (RR 0.23 [0.09 to 0.57]) (Appendix J Figure 53). Given the RD (RD -0.309 [-0.535 to -0.083]), for every four patients treated, one less patient would develop somnolence from treatment with levetiracetam than with carbamazepine.

One observational study comparing topiramate to carbamazepine reported data on somnolence.105 Risk of somnolence is nonsignificantly decreased by 72 percent when topiramate is used versus carbamazepine (RR 0.28 [0.02 to 3.38]) (Appendix J Figure 53).

One observational study comparing vigabatrin to carbamazepine reported data on somnolence.79 Risk of somnolence is nonsignificantly decreased by 46 percent when vigabatrin is used versus carbamazepine (RR 0.54 [0.12 to 2.13]) (Appendix J Figure 53).

Four observational studies comparing newer antiepileptic medications to carbamazepine reported data on somnolence.79,99,103,105 Risk of somnolence is significantly decreased by 70 percent when newer antiepileptic medications are used versus carbamazepine (RR 0.30 [0.14 to 0.62]) (Appendix J Figure 53). Given the RD (RD -0.044 [-0.117 to 0.029]), for every 23 patients treated, 1 less patient would develop somnolence from treatment with newer AED than with carbamazepine. No statistical heterogeneity (I2: 0 percent) or publication bias was detected (Egger's p=0.409).

One randomized controlled trial reported data on somnolence while patients were receiving levetiracetam compared with controlled release carbamazepine.62 Risk of somnolence is nonsignificantly increased by 21 percent when levetiracetam is used versus carbamazepine-CR (RR 1.21 [0.75 to 1.96]).

Phenytoin Versus Newer

Four trials reported somnolence while patients were receiving a newer antiepileptic medications compared with phenytoin57,59,64,90 and were all amenable to pooling.

One trial comparing lamotrigine to phenytoin reported data on somnolence.64 Risk of somnolence is significantly decreased by 75 percent when lamotrigine is used versus phenytoin (RR 0.25 [0.11 to 0.54]) (Appendix J Figure 54). Given the RD (RD -0.21 [-0.32 to -0.11]), for every five patients treated with lamotrigine, one less patient would develop somnolence compared with those treated with phenytoin.

Two trials comparing oxcarbazepine to phenytoin reported data on somnolence.57,59 Risk of somnolence is nonsignificantly decreased by 4 percent when oxcarbazepine is used versus phenytoin (RR 0.96 [0.72 to 1.28]) (Appendix J Figure 54).

One trial comparing topiramate to phenytoin reported data on somnolence.90 Risk of somnolence is nonsignificantly decreased by 14 percent when topiramate is used versus phenytoin (RR 0.86 [0.46 to 1.59]) (Appendix J Figure 54).

Four trials comparing newer antiepileptic medications to phenytoin reported data on somnolence.57,59,64,90 Risk of somnolence is nonsignificantly decreased by 28 percent when newer antiepileptic medications are used versus phenytoin (RR 0.72 [0.44 to 1.18]) (Appendix J Figure 54). A high level of statistical heterogeneity was detected (I2: 70.1 percent), but publication bias was not detected (Egger's p=0.124).

Valproic Acid Versus Newer

Eleven studies (nine randomized controlled trials and two observational studies) reported somnolence while patients were receiving a newer antiepileptic medications compared with valproic acid.50,51,58,68,72,76,77,84,89,99,105

Nine randomized controlled trials reported data on somnolence when newer antiepileptics were compared versus valproic acid.50,51,58,68,72,76,77,84,89 and were all amenable to pooling.

One trial comparing felbamate to valproic acid reported data on somnolence.51 Risk of somnolence is nonsignificantly decreased by 49 percent when felbamate is used versus valproic acid (RR 0.51 [0.07 to 3.78]) (Appendix J Figure 55).

Five trials comparing lamotrigine to valproic acid reported data on somnolence.50,68,72,76,89 Risk of somnolence is nonsignificantly increased by 27 percent when lamotrigine is used versus valproic acid (RR 0.73 [0.36 to 1.45]) (Appendix J Figure 55). A low level of statistical heterogeneity was detected (I2: 41.3 percent), but publication bias was not detected (Egger's p=0.869).

One trial comparing oxcarbazepine to valproic acid reported data on somnolence.58 Risk of somnolence is nonsignificantly decreased by 25 percent when oxcarbazepine is used versus valproic acid (RR 0.75 [0.43 to 1.28]) (Appendix J Figure 55).

Two trials comparing topiramate to valproic acid reported data on somnolence.77,84 Risk of somnolence is nonsignificantly decreased by 62 percent when topiramate is used versus valproic acid (RR 0.38 [0.12 to 1.21]) (Appendix J Figure 55). A low level of statistical heterogeneity was detected (I2: 25.7 percent), but tests for publication bias could not be performed.

Nine trials comparing newer antiepileptic medications to valproic acid reported data on somnolence.50,51,58,68,72,76,77,84,89 Risk of somnolence is significantly decreased by 35 percent when newer antiepileptic medications are used versus valproic acid (RR 0.65 [0.43 to 0.98]) (Appendix J Figure 55). Given the RD (RD -0.041 [-0.092 to 0.010]), for every 25 patients treated, 1 less patient would develop somnolence from treatment with newer antiepileptic medications than with valproic acid. A low level of statistical heterogeneity was detected (I2: 19.9 percent), but publication bias was not detected (Egger's p=0.639).

Two observational studies reported on somnolence while patients were receiving a newer antiepileptic medication compared with valproic acid and both were amenable for pooling.99,105

One observational study comparing lamotrigine to valproic acid reported data on somnolence.99 Risk of somnolence is nonsignificantly increased by 29 percent when lamotrigine is used versus valproic acid (RR 1.29 [0.14 to 12.28]) (Appendix J Figure 56).

One observational study comparing topiramate to valproic acid reported data on somnolence.105 Risk of somnolence is nonsignificantly decreased by 77 percent when topiramate is used versus valproic acid (RR 0.23 [0.02 to 2.46]) (Appendix J Figure 56).

Two observational studies comparing newer antiepileptic medications to valproic acid reported data on somnolence in patients with new onset epilepsy.99,105 Risk of somnolence is nonsignificantly decreased by 42 percent when newer antiepileptic medications are used versus valproic acid (RR 0.58 [0.08 to 4.38]) (Appendix J Figure 56).

Ethosuximide Versus Newer

One randomized controlled trial reported data on somnolence while patients were receiving lamotrigine compared with ethosuximide.89 The risk of somnolence was significantly decreased by 78 percent when lamotrigine was compared versus ethosuximide (RR 0.22 [0.07 to 0.70]). Given the RD (-0.07 [-0.12 to -0.020]), for every 15 patients treated with lamotrigine, 1 less patient would develop somnolence compared with those treated with ethosuximide.

Innovator Versus Generic Drug Evaluation

Somnolence is nonsignificantly decreased by 10 percent when generic carbamazepine was used versus innovator carbamazepine (RR 0.90 [0.48 to 1.70]). Analysis for any antiepileptic medications and BCS Class II antiepileptic medications is the same as carbamazepine analysis. No data was available to perform BCS Class I antiepileptic medications analysis.

Dizziness
Older Versus Newer Antiepileptic Drug Evaluation
Carbamazepine Versus Newer

Nineteen studies (16 randomized controlled trials and 3 observational studies) reported data on dizziness while patients were receiving a newer antiepileptic medications compared with carbamazepine.52,53,55,56,60,62,63,70,74,76-78,82,83,85,90,99,103,105

Sixteen randomized controlled trials reported data on dizziness when newer antiepileptics were compared versus carbamazepine52,53,55,56,60,62,63,70,74,76-78,82,83,85,90 and were all amenable to pooling.

Three trials comparing gabapentin to carbamazepine reported data on dizziness.60,78,85 Risk of dizziness is nonsignificantly decreased by 2 percent when gabapentin is used versus carbamazepine (RR 0.98 [0.75 to 1.28]) (Appendix J Figure 57). No statistical heterogeneity (I2: 0 percent) or publication bias was detected (Egger's p=0.785).

Seven trials comparing lamotrigine to carbamazepine reported data on dizziness.52,55,62,70,76,78,85 Risk of dizziness is significantly decreased by 21 percent when lamotrigine is used versus carbamazepine (RR 0.79 [0.64 to 0.97]) (Appendix J Figure 57). Given the RD (RD -0.015 [-0.034 to 0.005]), for every 67 patients treated, 1 less patient would develop dizziness from treatment with lamotrigine than with carbamazepine. No statistical heterogeneity (I2: 0 percent) or publication bias was detected (Egger's p=0.085).

Two trials comparing oxcarbazepine to carbamazepine reported data on dizziness.82,85 Risk of dizziness is nonsignificantly increased by 78 percent when oxcarbazepine is used versus carbamazepine (RR 1.78 [0.87 to 3.63]) (Appendix J Figure 57).

Five trials comparing topiramate to carbamazepine reported data on dizziness.74,77,83,85,90 Risk of dizziness is significantly decreased by 34 percent when topiramate is used versus carbamazepine (RR 0.66 [0.49 to 0.90]) (Appendix J Figure 57). Given the RD (RD -0.032 [-0.073 to 0.008]), for every 32 patients treated, 1 less patient would develop vomiting from treatment with topiramate than with carbamazepine. No significant statistical heterogeneity (I2: 0 percent) or publication bias was detected (Egger's p=0.292).

Three trials comparing vigabatrin to carbamazepine reported data on dizziness.53,56,63 Risk of dizziness is nonsignificantly decreased by 56 percent when vigabatrin is used versus carbamazepine (RR 0.44 [0.13 to 1.50]) (Appendix J Figure 57). A high level of statistical heterogeneity was detected (I2: 70.4 percent), but tests for publication bias could not be performed.

Sixteen trials comparing newer antiepileptic medications to carbamazepine reported data on dizziness.52,53,55,56,60,62,63,70,74,76-78,82,83,85,90 Risk of dizziness is significantly decreased by 22 percent when newer antiepileptic medications are used versus carbamazepine (RR 0.78 [0.67 to 0.91]) (Appendix J Figure 57). Given the RD (RD -0.020 [-0.041 to -0.0001]), for every 50 patients treated, 1 less patient would develop dizziness from treatment with newer AED than with carbamazepine. No statistical heterogeneity (I2: 0 percent) or publication bias was detected (Egger's p=0.256).

Three observational studies reported dizziness while patients were receiving a newer antiepileptic medication compared with carbamazepine and all were amenable for pooling.99,103,105

One observational study comparing lamotrigine to carbamazepine reported data on dizziness.99 Risk of dizziness is nonsignificantly decreased by 87 percent when lamotrigine is used versus carbamazepine (RR 0.13 [0.01 to 1.24]) (Appendix J Figure 58).

One observational study comparing topiramate to carbamazepine reported data on dizziness.105 Risk of dizziness is nonsignificantly decreased by 58 percent when topiramate is used versus carbamazepine (RR 0.42 [0.06 to 3.17])(Appendix J Figure 58).

Three observational studies comparing newer antiepileptic medications to carbamazepine reported data on dizziness.99,103,105 Risk of dizziness is significantly decreased by 79 percent when newer antiepileptic medications are used versus carbamazepine (RR 0.21 [0.05 to 0.89]) (Appendix J Figure 58). Given the RD (RD -0.028 [-0.062 to 0.005]), for every 36 patients treated, 1 less patient would develop dizziness from treatment with newer AED than with carbamazepine. No statistical heterogeneity was detected (I2: 0 percent), but tests for publication bias could not be performed.

One observational study comparing newer antiepileptic medications with carbamazepine reported data on dizziness in patients with new partial epilepsy.103 Risk of dizziness is nonsignificantly decreased by 85 percent when levetiracetam is used versus carbamazepine (RR 0.15 [0.02 to 1.13]) (Appendix J Figure 58).

Two randomized controlled trials reported data on dizziness when newer antiepileptics were compared with controlled- or sustained-release carbamazepine and both were amenable for pooling.81,86

One randomized controlled trial reported data on dizziness when lamotrigine was compared with sustained-release carbamazepine.86 The risk of dizziness was nonsignificantly increased by 43 percent when lamotrigine was used versus sustained release carbamazepine (RR 1.43 [0.66 to 3.13]) (Appendix J Figure 59).

One randomized controlled trial reported data on dizziness when levetiracetam was compared with controlled-release carbamazepine.81 The risk of dizziness was nonsignificantly decreased by 21 percent when levetiracetam is used versus carbamazepine-CR (RR 0.79 [0.51 to 1.22]) (Appendix J Figure 59).

Two randomized controlled trials reported data on dizziness when lamotrigine or levetiracetam were compared with controlled- or sustained-release carbamazepine and both were amenable for pooling.81,86 The risk of dizziness was nonsignificantly decreased by 4 percent when either newer agent was compared with controlled- or sustained-release carbamazepine (RR 0.96 [0.56 to 1.66]) (Appendix J Figure 59).

Phenytoin Versus Newer

Three randomized controlled trials reported dizziness while patients were receiving newer antiepileptic medications compared with phenytoin57,59,64 and were all amenable to pooling.

One trial comparing lamotrigine to phenytoin reported data on dizziness.64 Risk of dizziness is nonsignificantly decreased by 20 percent when lamotrigine is used versus phenytoin (RR 0.80 [0.35 to 1.85]) (Appendix J Figure 60).

Two trials comparing oxcarbazepine to phenytoin reported data on dizziness.57,59 Risk of dizziness is nonsignificantly decreased by 38 percent when oxcarbazepine is used versus phenytoin (RR 0.62 [0.31 to 1.24]) (Appendix J Figure 60).

Three trials comparing newer antiepileptic medications to phenytoin reported data on dizziness.57,59,64 Risk of dizziness is nonsignificantly decreased by 33 percent when newer antiepileptic medications are used versus phenytoin (RR 0.67 [0.43 to 1.05]) (Appendix J Figure 60). No statistical heterogeneity was detected (I2: 18.1 percent) but tests for publication bias could not be performed.

One observational study comparing newer antiepileptic medications to phenytoin reported data on dizziness in children 18 years of age or younger.59 Risk of dizziness is significantly decreased by 58 percent when newer antiepileptic medications are used versus phenytoin (RR 0.42 [0.20 to 0.85]). Given the RD (RD -0.130 [-0.232 to -0.027]), for every eight patients treated, one less patient would develop dizziness from treatment with newer AED than with phenytoin.

Valproic Acid Versus Newer

Fourteen studies (12 randomized controlled trials and 2 observational studies) reported dizziness while patients were receiving a newer antiepileptic medications compared with valproic acid.51,58,68,72,74,76,77,82,84,85,88,89,99,105

Twelve randomized controlled trials reported data on dizziness when newer antiepileptics were compared versus valproic acid51,58,68,72,74,76,77,82,84,85,88,89 and were all amenable to pooling.

One trial comparing felbamate to valproic acid reported data on dizziness.51 Risk of dizziness is nonsignificantly decreased by 49 percent when felbamate is used versus valproic acid (RR 0.51 [0.07 to 3.78]) (Appendix J Figure 61).

Six trials comparing lamotrigine to valproic acid reported data on dizziness.68,72,76,85,88,89 Risk of dizziness is nonsignificantly increased by 34 percent when lamotrigine is used versus valproic acid (RR 1.34 [0.85 to 2.12]) (Appendix J Figure 61). No statistical heterogeneity was detected (I2: 0 percent), but statistically significant publication bias was detected (Egger's p=0.006).

Two trials comparing oxcarbazepine to valproic acid reported data on dizziness.58,82 Risk of dizziness is nonsignificantly increased by 15 percent when oxcarbazepine is used versus valproic acid (RR 1.15 [0.33 to 3.97]) (Appendix J Figure 61).

Four trials comparing topiramate to valproic acid reported data on dizziness.74,77,84,85 Risk of dizziness is nonsignificantly decreased by 27 percent when topiramate is used versus valproic acid (RR 0.73 [0.36 to 1.49]) (Appendix J Figure 61). A low level of statistical heterogeneity (I2: 16.3 percent) was detected, but no publication bias was detected (Egger's p=0.113).

Twelve trials comparing newer antiepileptic medications to valproic acid reported data on dizziness.51,58,68,72,74,76,77,82,84,85,88,89 Risk of dizziness is nonsignificantly decreased by 2 percent when newer antiepileptic medications are used versus valproic acid (RR 0.98 [0.71 to 1.35]) (Appendix J Figure 61). No statistical heterogeneity (I2: 0 percent) or publication bias was detected (Egger's p=0.111).

Two observational studies reported dizziness while patients were receiving a newer antiepileptic medication compared with carbamazepine and both were amenable for pooling.99,105

One observational study comparing lamotrigine to valproic acid reported data on dizziness.99 Risk of dizziness is nonsignificantly decreased by 57 percent when lamotrigine is used versus valproic acid (RR 0.43 [0.04 to 5.17]) (Appendix J Figure 62).

One observational study comparing topiramate to valproic acid reported data on dizziness.105 Risk of dizziness is nonsignificantly decreased by 45 percent when topiramate is used versus valproic acid (RR 0.55 [0.08 to 3.76]) (Appendix J Figure 62).

Two observational studies comparing newer antiepileptic medications to valproic acid reported data on dizziness.99,105 Risk of dizziness is nonsignificantly decreased by 50 percent when newer antiepileptic medications are used versus valproic acid (RR 0.50 [0.08 to 3.18]) (Appendix J Figure 62).

Ethosuximide Versus Newer

One randomized controlled trial reported data on dizziness while patients were receiving lamotrigine compared with ethosuximide.89 The risk of dizziness was nonsignificantly decreased by 54 percent when lamotrigine was compared versus ethosuximide (RR 0.46 [0.15 to 1.38]).

Innovator Versus Generic Drug Evaluation

Dizziness is nonsignificantly decreased by 50 percent when generic carbamazepine was used versus innovator carbamazepine (RR 0.50 [0.07 to 3.33]) in the one trial that reported this.109 Results for any antiepileptic medications and BCS Class II antiepileptic medications is the same as carbamazepine results. No data was available for BCS Class I antiepileptic medications.

Combined Neurological Events and Neurological Components
Innovator Versus Generic Drug Evaluation

One controlled clinical trial110 and one controlled observational study130reported data on combined neurological adverse events. Results from the clinical trial showed that 5 out of 23 patients experienced neurological adverse events while receiving generic carbamazepine while only 1 out of 23 patients experienced neurological adverse event while on innovator carbamazepine. This difference was not statistically significant. Data from the observational trial is reported as events per 1,000 person-years (percent) for innovator and generic carbamazepine. In the generic group (n=705), there were 145.7 events per 1,000 person-years (percent), whereas in the innovator group (n=275), there were 75.7 events per 1,000 person-years (percent). This difference is also not statistically significant. None of the trials used discernable FDA “A” rated generics.

Results for any antiepileptic medications and BCS Class II antiepileptic drugs are the same as carbamazepine results given above. Data on BCS Class I antiepileptic medications were not available.

Three controlled clinical trials109,110,119 and no controlled observational studies reported individual neurological adverse events. Headache was reported by all three trials, diplopia and somnolence was reported by two trials,109,110 and dizziness was reported by only one trial.109 None of the trials used discernable FDA “A” rated generics.

Diplopia was nonsignificantly increased by 28 percent when generic carbamazepine was used versus innovator carbamazepine (RR 1.28 [0.38 to 4.31]). Analysis for any antiepileptic medications and BCS Class II antiepileptic medications is the same as carbamazepine analysis. No data was available to perform BCS Class I antiepileptic medications analysis.

No controlled clinical trials or controlled observational studies reported data on asthenia, ataxia, nystagmus, or tremor.

Hypotension
Innovator Versus Generic Drug Evaluation

No controlled clinical trials or observational studies comparing innovator versus generic antiepileptic medications reported data on hypotension.

Nausea
Older Versus Newer Antiepileptic Drug Evaluation
Carbamazepine Versus Newer

Eight randomized controlled trials reported data on nausea when newer antiepileptics were compared versus carbamazepine47,52,55,56,74,76,77,85 and were all amenable to pooling.

One trial comparing gabapentin to carbamazepine reported data on nausea.85 Risk of nausea is nonsignificantly decreased by 22 percent when gabapentin is used versus carbamazepine (RR 0.78 [0.30 to 2.00]) (Appendix J Figure 63).

Four trials comparing lamotrigine to carbamazepine reported data on nausea.52,55,76,85 Risk of nausea is nonsignificantly decreased by 5 percent when lamotrigine is used versus carbamazepine (RR 0.95 [0.56 to 1.60]) (Appendix J Figure 63). A low level of statistical heterogeneity was detected (I2: 25.9 percent), but significant publication bias was detected (Egger's p=0.005).

Two trials comparing oxcarbazepine to carbamazepine reported data on nausea.47,85 Risk of nausea is nonsignificantly increased by 87 percent when oxcarbazepine is used versus carbamazepine (RR 1.87 [0.34 to 10.40]) (Appendix J Figure 63).

Three trials comparing topiramate to carbamazepine reported data on nausea.74,77,85 Risk of nausea is significantly decreased by 51 percent when topiramate is used versus carbamazepine (RR 0.49 [0.33 to 0.74]) (Appendix J Figure 63). Given the RD (RD -0.073 [-0.166 to 0.019]), for every 14 patients treated, 1 less patient would develop nausea from treatment with topiramate than with carbamazepine. No significant statistical heterogeneity (I2: 0 percent) and no publication bias was detected (Egger's p=0.300).

One trial comparing vigabatrin to carbamazepine reported data on nausea.56 Risk of nausea is nonsignificantly decreased by 76 percent when vigabatrin is used versus carbamazepine (RR 0.24 [0.04 to 1.47]) (Appendix J Figure 63).

Eight trials comparing newer antiepileptic medications to carbamazepine reported data on nausea.47,52,55,56,74,76,77,85 Risk of nausea is nonsignificantly decreased by 31 percent when newer antiepileptic medications are used versus carbamazepine (RR 0.69 [0.46 to 1.02]) (Appendix J Figure 63). A low level of statistical heterogeneity was detected (I2: 32.7 percent), but no publication bias were detected (Egger's p=0.218).

One randomized controlled trial reported data on nausea while patients were receiving levetiracetam compared with controlled-release carbamazepine.81 Risk of nausea is nonsignificantly decreased by 34 percent when levetiracetam is used versus controlled-release carbamazepine (RR 0.66 [0.39 to 1.12]).

Phenytoin Versus Newer

Four randomized controlled trials reported nausea while patients were receiving a newer antiepileptic medication compared with phenytoin57,59,64,90 and were all amenable to pooling.

One trial comparing lamotrigine to phenytoin reported data on nausea.64 Risk of nausea is nonsignificantly increased by 93 percent when lamotrigine is used versus phenytoin (RR 1.93 [0.63 to 6.02]) (Appendix J Figure 64).

Two trials comparing oxcarbazepine to phenytoin reported data on nausea.57,59 Risk of nausea is nonsignificantly decreased by 20 percent when oxcarbazepine is used versus phenytoin (RR 0.80 [0.45 to 1.45]) (Appendix J Figure 64).

One trial comparing topiramate to phenytoin reported data on nausea.90 Risk of nausea is nonsignificantly decreased by 28 percent when topiramate is used versus phenytoin (RR 0.72 [0.32 to 1.62]) (Appendix J Figure 64).

Four trials comparing newer antiepileptic medications to phenytoin reported data on nausea.57,59,64,90 Risk of nausea is nonsignificantly decreased by 12 percent when newer antiepileptic medications are used versus phenytoin (RR 0.88 [0.56 to 1.37]) (Appendix J Figure 64). No statistical heterogeneity (I2: 0 percent) or publication bias was detected (Egger's p=0.500).

Valproic Acid Versus Newer

Eleven randomized controlled trials reported data on nausea when newer antiepileptics were compared versus valproic acid51,58,68,72,74,76,77,84,85,88,89 and were all amenable to pooling.

One trial comparing felbamate to valproic acid reported data on nausea.51 Risk of nausea is nonsignificantly decreased by 66 percent when felbamate is used versus valproic acid (RR 0.34 [0.03 to 4.04]) (Appendix J Figure 65).

Six trials comparing lamotrigine to valproic acid reported data on nausea.68,72,76,85,88,89 Risk of nausea is significantly decreased by 52 percent when lamotrigine is used versus valproic acid (RR 0.48 [0.27 to 0.86]) (Appendix J Figure 65). Given the RD (RD -0.043 [-0.086 to -0.001]), for every 24 patients treated, 1 less patient would develop nausea from treatment with lamotrigine than with valproic acid. A low level of statistical heterogeneity was detected (I2: 33.8 percent), but publication bias was not detected (Egger's p=0.982).

One trial comparing oxcarbazepine to valproic acid reported data on nausea.58 Risk of nausea is nonsignificantly decreased by 26 percent when oxcarbazepine is used versus valproic acid (RR 0.74 [0.36 to 1.54]) (Appendix J Figure 65).

Four trials comparing topiramate to valproic acid reported data on nausea.74,77,84,85 Risk of nausea is nonsignificantly decreased by 39 percent when topiramate is used versus valproic acid (RR 0.61 [0.36 to 1.04]) (Appendix J Figure 65). No significant statistical heterogeneity (I2: 0 percent) or publication bias was detected (Egger's p=0.170).

Eleven trials comparing newer antiepileptic medications to valproic acid reported data on nausea.51,58,68,72,74,76,77,84,85,88,89 Risk of nausea is significantly decreased by 44 percent when newer antiepileptic medications are used versus valproic acid (RR 0.56 [0.41 to 0.77]) (Appendix J Figure 65). Given the RD (RD -0.033 [-0.056 to -0.010]), for every 31 patients treated, 1 less patient would develop nausea from treatment with newer AED than with valproic acid. No statistical heterogeneity (I2: 0 percent) or publication bias was detected (Egger's p=0.272).

Innovator Versus Generic Drug Evaluation

No controlled clinical trials or observational studies comparing innovator versus generic antiepileptic medications reported data on nausea.

Vomiting
Older Versus Newer Antiepileptic Drug Evaluation
Carbamazepine Versus Newer

Three randomized controlled trials reported data on vomiting when newer antiepileptics were compared versus carbamazepine.52,62,77 and were all amenable to pooling.

Two trials comparing lamotrigine to carbamazepine reported data on vomiting.52,62 Risk of vomiting is nonsignificantly increased by 34 percent when lamotrigine is used versus carbamazepine (RR 1.34 [0.67 to 2.68]) (Appendix J Figure 66).

One trial comparing topiramate to carbamazepine reported data on vomiting.77 Risk of vomiting is nonsignificantly decreased by 12 percent when topiramate is used versus carbamazepine (RR 0.88 [0.19 to 4.12]) (Appendix J Figure 66).

Three trials comparing newer antiepileptic medications to carbamazepine reported data on vomiting.52,62,77 Risk of vomiting is nonsignificantly increased by 25 percent when newer antiepileptic medications are used versus carbamazepine (RR 1.25 [0.66 to 2.35]) (Appendix J Figure 66). No statistical heterogeneity (I2: 0 percent) or publication bias was detected (Egger's p=0.315).

Three trials comparing newer antiepileptic medications to carbamazepine reported data on vomiting in patients with new onset epilepsy.52,62,77 Risk of vomiting is nonsignificantly increased by 25 percent when newer antiepileptic medications are used versus carbamazepine (RR 1.25 [0.66 to 2.35]) (Appendix J Figure 66). No statistical heterogeneity (I2: 0 percent) or publication bias was detected (Egger's p=0.315).

Phenytoin Versus Newer

Only one randomized controlled trial reported data on vomiting while patients were receiving a newer antiepileptic medication (oxcarbazepine) compared with phenytoin.59 This trial was in children 18 years of age or younger. Risk of vomiting is significantly decreased by 91 percent when oxcarbazepine is used versus phenytoin (RR 0.09 [0.01 to 0.89]) (Appendix J Figure 67). Given the RD (RD -0.053 [-0.102 to -0.004]), for every 19 patients treated, 1 less patient would develop vomiting from treatment with oxcarbazepine than with phenytoin.

Valproic Acid Versus Newer

Five randomized controlled trials reported data on vomiting while patients were receiving a newer antiepileptic medications compared with valproic acid.51,72,72,77,88

One trial comparing felbamate to valproic acid reported data on vomiting.76 Risk of vomiting is nonsignificantly increased by 3 percent when felbamate is used versus valproic acid (RR 1.03 [0.19 to 5.61]) (Appendix J Figure 68).

Three trials comparing lamotrigine to valproic acid reported data on vomiting.68,72,88 Risk of vomiting is nonsignificantly decreased by 34 percent when lamotrigine is used versus valproic acid (RR 0.66 [0.22 to 2.00]) (Appendix J Figure 68). A low level of statistical heterogeneity was detected (I2: 55.9 percent), but tests for publication bias could not be performed.

One trial comparing topiramate to valproic acid reported data on vomiting.77 Risk of vomiting is nonsignificantly increased by 65 percent when topiramate is used versus valproic acid (RR 1.65 [0.21 to 12.79]) (Appendix J Figure 68).

Five trials comparing newer antiepileptic medications to valproic acid reported data on vomiting.51,68,72,77,88 Risk of vomiting is nonsignificantly decreased by 31 percent when newer antiepileptic medications are used versus valproic acid (RR 0.69 [0.34 to 1.42]) (Appendix J Figure 68). A low level of statistical heterogeneity was detected (I2: 16.5 percent) and statistically significant publication bias was detected (Egger's p=0.036).

Innovator Versus Generic Drug Evaluation

No controlled clinical trials or observational studies comparing innovator versus generic antiepileptic medications reported data on vomiting.

Skin Rash
Older Versus Newer Antiepileptic Drug Evaluation
Carbamazepine Versus Newer

Eighteen studies (13 randomized controlled trials and 5 observational studies) reported skin rash while patients were receiving a newer antiepileptic medications compared with carbamazepine.52,53,55,56,62,63,70,74,78,80,82,83,85,95,99,103-105

Thirteen randomized controlled trials reported data on skin rash when newer antiepileptics were compared versus carbamazepine52,53,55,56,62,63,70,74,78,80,82,83,85 and were all amenable to pooling.

Two randomized controlled trials comparing gabapentin to carbamazepine reported data on skin rash and were amenable for pooling.78,85 Risk of skin rash is significantly decreased by 69 percent when gabapentin is used versus carbamazepine (RR 0.31 [0.14 to 0.69]) (Appendix J Figure 69). Given the RD (RD -0.049 [-0.084 to -0.015]), for every 20 patients treated, 1 less patient would develop skin rash from treatment with gabapentin than with carbamazepine.

Seven randomized controlled trials comparing lamotrigine to carbamazepine reported data on skin rash.52,55,62,70,78,80,85 Risk of skin rash is significantly decreased by 34 percent when lamotrigine is used versus carbamazepine (RR 0.66 [0.46 to 0.94]) (Appendix J Figure 69). Given the RD (RD -0.032 [-0.053 to -0.012]), for every 32 patients treated, 1 less patient would develop skin rash from treatment with lamotrigine than with carbamazepine. A low level of statistical heterogeneity was detected (I2: 39.1 percent), but publication bias was not detected (Egger's p=0.234).

Two trials comparing oxcarbazepine to carbamazepine reported data on skin rash.82,85 Risk of skin rash is nonsignificantly decreased by 9 percent when oxcarbazepine is used versus carbamazepine (RR 0.91 [0.56 to 1.48]) (Appendix J Figure 69).

Three trials comparing topiramate to carbamazepine reported data on skin rash.74,83,85 Risk of skin rash is significantly decreased by 62 percent when topiramate is used versus carbamazepine (RR 0.38 [0.24 to 0.59]) (Appendix J Figure 69). Given the RD (RD -0.061 [-0.090 to -0.033]), for every 17 patients treated, 1 less patient would develop skin rash from treatment with topiramate than with carbamazepine. No significant statistical heterogeneity was detected (I2: 0 percent), but publication bias was detected (Egger's p=0.050).

Three trials comparing vigabatrin to carbamazepine reported data on skin rash.53,56,63 Risk of skin rash is significantly decreased by 71 percent when vigabatrin is used versus carbamazepine (RR 0.29 [0.13 to 0.62]) (Appendix J Figure 69). Given the RD (RD -0.075 [-0.125 to -0.026]), for every 14 patients treated, 1 less patient would develop skin rash from treatment with vigabatrin than with carbamazepine. No significant statistical heterogeneity was detected (I2: 0 percent) and tests for publication bias could not be performed.

Thirteen randomized controlled trials comparing newer antiepileptic medications to carbamazepine reported data on skin rash.52,53,55,56,62,63,70,74,78,80,82,83,85 Risk of skin rash is significantly decreased by 48 percent when newer antiepileptic medications are used versus carbamazepine (RR 0.52 [0.39 to 0.69]) (Appendix J Figure 69). Given the RD (RD -0.043 [-0.058 to -0.028]), for every 24 patients treated, 1 less patient would develop skin rash from treatment with newer AED than with carbamazepine. A low level of statistical heterogeneity was detected (I2: 35.2 percent) and significant publication bias was detected (Egger's p=0.001).

Five observational studies reported data on skin rash while patients were receiving a newer antiepileptic medication compared with carbamazepine and all five were amenable for pooling.95,99,103-105

One observational trial comparing lamotrigine to carbamazepine reported data on skin rash.99 Risk of skin rash is nonsignificantly decreased by 63 percent when lamotrigine is used versus carbamazepine (RR 0.37 [0.12 to 1.11]) (Appendix J Figure 70).

One observational study comparing levetiracetam to carbamazepine reported data on skin rash.103 Risk of skin rash is nonsignificantly decreased by 7 percent when levetiracetam is used versus carbamazepine (RR 0.93 [0.08 to 11.16]) (Appendix J Figure 70).

One observational study comparing oxcarbazepine to carbamazepine reported data on skin rash.95 Risk of skin rash is nonsignificantly increased 3-fold when oxcarbazepine is used versus carbamazepine (RR 3.00 [0.26 to 35.71]) (Appendix J Figure 70).

Two observational studies comparing topiramate to carbamazepine reported data on skin rash.104,105 Risk of skin rash is significantly decreased by 90 percent when topiramate is used versus carbamazepine (RR 0.10 [0.02 to 0.49]) (Appendix J Figure 70). Given the RD (RD -0.073 [-0.171 to 0.025]), for every 14 patients treated, 1 less patient would develop skin rash from treatment with topiramate than with carbamazepine.

Five observational studies comparing newer antiepileptic medications to carbamazepine reported data on skin rash.95,99,103-105 Risk of skin rash is nonsignificantly decreased by 67 percent when newer antiepileptic medications are used versus carbamazepine (RR 0.33 [0.10 to 1.05]) (Appendix J Figure 70). A low level of statistical heterogeneity was detected (I2: 28.1 percent), but publication bias was not detected (Egger's p=0.671).

Two randomized controlled trials reported data on skin rash when newer antiepileptics were compared with controlled- or sustained-release carbamazepine and both were amenable for pooling.81,86

One randomized controlled trial reported data on skin rash when lamotrigine was compared with sustained release carbamazepine.86 The risk of skin rash was nonsignificantly decreased by 59 percent when lamotrigine was used versus sustained release carbamazepine (RR 0.41 [0.16 to 1.07]) (Appendix J Figure 71).

One randomized controlled trial reported data on skin rash when levetiracetam was compared with controlled-release carbamazepine.81 The risk of skin rash was nonsignificantly decreased by 49 percent when levetiracetam is used versus controlled-release carbamazepine (RR 0.51 [0.23 to 1.15]) (Appendix J Figure 71).

Two randomized controlled trials reported data on skin rash when lamotrigine or levetiracetam was compared with controlled- or sustained-release carbamazepine and both were amenable for pooling.81,86 The risk of skin rash was significantly decreased by 53 percent when either newer agent was compared with controlled- or sustained-release carbamazepine (RR 0.47 [0.25 to 0.89]) (Appendix J Figure 71). Given the RD (RD [-0.038 [-0.081 to 0.005]), for every 27 patients treated with a newer agent, 1 less patient would develop skin rash compared with those treated with controlled- or sustained-release carbamazepine.

Phenytoin Versus Newer

Four randomized controlled trials reported skin rash while patients were receiving a newer antiepileptic medication compared with phenytoin57,59,64,90 and were all amenable to pooling.

One randomized controlled trial comparing lamotrigine to phenytoin reported data on skin rash.64 Risk of skin rash is nonsignificantly increased by 47 percent when lamotrigine is used versus phenytoin (RR 1.47 [0.6 to 3.26]) (Appendix J Figure 72).

Two trials comparing oxcarbazepine to phenytoin reported data on skin rash.57,59 Risk of skin rash is nonsignificantly decreased by 22 percent when oxcarbazepine is used versus phenytoin (RR 0.78 [0.42 to 1.46]) (Appendix J Figure 72).

One trial comparing topiramate to phenytoin reported data on skin rash.90 Risk of skin rash is significantly decreased by 90 percent when topiramate is used versus phenytoin (RR 0.10 [0.02 to 0.57]) (Appendix J Figure 72). Given the RD (RD -0.071 [-0.120 to -0.022]), for every 15 patients treated, 1 less patient would develop skin rash from treatment with topiramate than with phenytoin.

Four trials comparing newer antiepileptic medications to phenytoin reported data on skin rash.57,59,64,90 Risk of skin rash is nonsignificantly decreased by 24 percent when newer antiepileptic medications are used versus phenytoin (RR 0.76 [0.34 to 1.66]) (Appendix J Figure 72). A high level of statistical heterogeneity was detected (I2: 54 percent), but publication bias was not detected (Egger's p=0.256).

Valproic Acid Versus Newer

Twelve studies (10 randomized controlled trials and 2 observational studies) reported skin rash while patients were receiving a newer antiepileptic medications compared with valproic acid.51,72,74,75,80,82,84,85,88,94,99,105

Ten randomized controlled trials reported data on skin rash when newer antiepileptics were compared versus valproic acid.51,72,74,75,80,82,84,85,