This is the first known study to find that antiepileptic drug refilling itself may be associated with an elevated risk of seizure-related outcomes. We observed that, with dose and dosage form held constant, refilling the same antiepileptic from the same manufacture was associated with an approximately 2.1- to 2.3-fold increase in odds of emergency treatment related to seizure, and that points estimates for switching between antiepileptic drugs from different manufacturers were similar, but with wide confidence intervals. The refill-adjusted effect of switching between products from different manufacturers was small, with a 4% to 19% increase in odds of seizure-related outcomes.
While the exact mechanisms underlying these phenomena have not been determined, we propose two possible explanations. First, our results are consistent with hypotheses that suggest that variability in antiepileptic bioavailability can lead to a small increased risk of breakthrough seizures. A review of more than 2,000 clinical bioequivalence studies of orally administered generic products approved by FDA found that the average difference in Cmax
and AUC between generic and innovator products was 4.4% and 3.6%, respectively (19
). Given inherent variability in manufacturing processes, similar fluctuations in bioavailability between lots produced by the same manufacturer are expected (20
) and have been observed (23
A second possible explanation is that refilling behavior may result in lapses in pharmacotherapy continuity which may lead to breakthrough seizures. The prescription refilling process (and switching process, to the extent that switching is a special case of refilling) involves many time-sensitive steps, often beginning with a visit or telephone call to a prescribing physician or to a dispensing pharmacy, to picking up the medication from the dispensing pharmacy, to beginning the new refill on the right day. A failure or delay at any point along this chain of events could result in a brief lapse in antiepileptic drug adherence, which may cause breakthrough seizures (24
). Labels on newly dispensed prescriptions may also differ from the previous dispensing, and confusion about dosing or administration may result (26
). Between-manufacturer variation in peripheral features of the drug product may further contribute to such behavioral explanations. For example, patients stable on particular drug products of given size, shape, and color, may regard new prescriptions that vary in one or more of these features with caution, perhaps going so far as to not take the medication for fear of having received the wrong drug from the pharmacy, thus increasing the likelihood of lapses in pharmacotherapy continuity. Some have suggested that variation in drug use patterns is at least as important as pharmacokinetics in explaining variation in drug response (10
Regardless of the mechanism, we observed that both refilling and switching of antiepileptic medications were associated with an elevated risk of seizure-related events. While this is an important finding, the refill-adjusted analysis allowed us to ascertain the extent to which switching between antiepileptic drugs is associated with seizure-related outcomes after accounting for within-manufacturer variability and aspects of patient behavior associated with breakthrough seizure inherent in the prescription refilling/switching process.
The results of our refill-adjusted analyses are consistent with a recent meta-analysis of bioequivalence trials that reported seizure outcomes (17
) and with a recent case-control study (16
). Other case-control studies suggest that the increased risk of seizure-related outcomes associated with switching versus not switching ranges between 78% and 84% (13
). However, these studies did not consider the effect of within-manufacturer variability in bioavailability or other factors inherent in the refilling and switching process, which likely inflated their findings. Furthermore, confounding by epilepsy severity may limit these studies since epilepsy that is not well controlled would predispose to seizure and would also be associated with use of multiple antiepileptic medications, thus increasing the probability that there would be at least one medication switch. Hansen et al partly adjusted for this confounding bias by adjusting for number of antiepileptic medications dispensed, which reduced the primary effect from 1.78 (95% CI, 1.35–2.36) to 1.57 (95% CI, 1.17–2.10) and indicated number of antiepileptic medications was a strong predictor of seizure-related events (15
). Indeed, the one case-control study that was consistent with our findings adjusted for several potential confounders, including total number of antiepileptic medications and use of interacting medications, although conditioning on intermediates may be a concern in that study (16
). The case-crossover approach is valid regardless of the number of drugs used since the frequency of refills and switches would be expected to be uniformly distributed across case- and control-periods of short duration, in the absence of a true effect and of bias.
To assess the validity of the duration of the pre-defined case- and control-periods, we plotted the distributions of exposures (both refills and switches) over the 8 weeks immediately preceding the index event for each patient. Both refills and switches were largely clustered within the 21 days preceding the index events, substantiating the use of the 21-day periods and suggesting that the risk of seizure-related events may be greatest in the 3 weeks following an antiepileptic refill or switch. Furthermore, case periods of greater than 21 days would lead to misclassification of the exposure. This explains why the effect estimates for analyses using 28-day periods are smaller than those using 21-day periods.
Several important limitations of this study should be noted. First, we identified seizures using inpatient and ER ICD codes. The accuracy of ICD codes for identifying seizure is not known and many patients do not seek emergency treatment for seizure so our study population represents a select sample of patients who experience a seizure of severity great enough to warrant hospitalization. While this highly specific outcome definition may limit the generalizability of the results, it preserves the internal validity of ratio effect measures (29
). We likely further improved the specificity of the outcome by requiring current antiepileptic drug use and an outpatient diagnosis of epilepsy in the year prior to the outcome.
A second limitation is that, should a bioavailability mechanism be implicated, focusing on seizure-related outcomes ignores the other end of the adverse event spectrum – namely, toxicities resulting from exceeding the upper limit of the therapeutic window for drug plasma levels. Nevertheless, such adverse events would be difficult to capture in administrative data and are less important than breakthrough seizures given the clinical ramifications of seizures (30
). Another limitation of this analysis is the small number of cases on which it is based, particularly in subgroups of refilling or switching type. However, because seizure-related emergency treatment is fairly rare and because only those with an exposure in either the case- or the control-period, but not both, contribute to the analyses of case-crossover studies, this represents a rare disease – rare exposure scenario. That a sizeable proportion of prescription dispensings in BC are for 90-days supply, rather than 30 days, further reduces the total number of possible exposures.
We expect that the implications of this study can be extended beyond Canada and that the association between antiepileptic switching and seizure-related events may be smaller in other countries since EMA and US FDA bioequivalence requirements are more stringent than those of Health Canada (1
). Requirements established by FDA, which are identical to those of EMA, are intended to ensure that differences in bioavailability between bioequivalent products are no greater than between-lot variation from a single manufacturer (31
). Indeed, this study is the first to describe an association between refilling the same antiepileptic medication and seizure-related events and we also found that this association is of similar magnitude as that for prescription switching. Nevertheless, some advocacy groups (32
), professional organizations (6
), and legislators (33
) oppose antiepileptic drug substitution and at least one US state prevents substitution by pharmacists (34
While our results do not completely rule out the possibility that switching between antiepileptic medications produced by different manufacturers may contribute to loss of seizure control in some patients, our findings indicate that after adjustment for the risk associated with refilling the same agent, the residual harmful effect of switching between two generic formulations of the same medication or between a brand and a generic (or vice versa) of the same drug is either negligible or much smaller than that reported in earlier studies that did not adjust properly for the effect of prescription refilling per se.