The total defined daily dose of antiepileptic drugs was the best predictor of medication side effects in a clinical prediction rule developed from a single-center, cross-sectional study of 801 patients with epilepsy.

The finding stands in contrast to a previous study that found overall antiepileptic drug (AED) burden, also calculated by defined daily dose (DDD), was not a significant predictor of the risk of side effects (Epilepsia 2010;51:797-804).

But the lead author of the current study, Jonathan Dykeman, and his associates used recursive partitioning to develop a clinical decision model that classifies patients into subgroups based on their risk of side effects, whereas the earlier report used standard multivariate linear progression to model patients’ scores on the Adverse Event Profile questionnaire. Disentangling the number of AEDs a person used from the overall AED burden on standard linear regression is not easy because the two are highly correlated, he said.

The investigators aim to help guide therapy and clinical decision making with their clinical prediction rule. “To our knowledge, there isn’t a clinical prediction rule that’s used because a lot of the focus is on reduction of seizures,” said Mr. Dykeman, an MD/PhD student in the department of clinical neurosciences at the University of Calgary (Alta.) at the annual meeting of the American Academy of Neurology.

Drug burden was high in the group of 801 patients; they took 17 different AEDs, and there were 132 possible combinations of drugs. Instead of modeling that complex scenario, Mr. Dykeman and his colleagues converted AED burden into the World Health Organization’s definition of DDD, which is the average total that a person would take in 1 day for the drug’s main indication. The investigators also included a number of other characteristics that broadly fit into the categories of AED type, sociodemographic factors, and clinical factors. Patients reported side effects during clinical interviews.

The recursive partitioning method hierarchically organized the risk factors associated with side effects from strongest to weakest. Overall, 18% of the cohort experienced side effects. A DDD of greater than 3.5 was the best predictor of side effects. Of 46 patients with a DDD greater than 3.5, 33% reported side effects, compared with 17% of 755 patients with a lower DDD.

A history of psychiatric treatment with medication or counseling also carried about a threefold higher risk for side effects than did the absence of such history. Among patients with the higher DDD, side effects occurred in 8 (57%) of 14 patients with a history of psychiatric treatment, compared with 22% of low DDD patients without a history of psychiatric treatment.

“We have a side study going that’s going to try to tease out whether that’s an interaction with the psychiatric drugs or the fact that they have a psychiatric illness,” Mr. Dykeman said in an interview.

Also, side effects reportedly occurred significantly less often among patients on a low DDD with a history of a learning disability (9% of 111) than in patients without such history (18% of 644). “We highly suspect this is related more to the reporting of side effects than actually the occurrence of side effects,” said Mr. Dykeman. It might have been more difficult to convey that the patients were having side effects because of their learning disorders as opposed to not actually having them, he said.

The group of patients without learning disabilities could be further partitioned for side-effect risk based on whether they used topiramate (35% of 37) or not (17% of 607). Age was used to further refine the risk for side effects among patients who did not take topiramate. Patients aged 45 years or younger (6 of 27, or 22%) had a lower risk for side effects, compared with patients older than 45 years (7 of 10, or 70%).

Mr. Dykeman said the conclusions that can be drawn from the study are limited by the lack of ascertainment of self-reported side effects, the cross-sectional nature of the data, and the small sizes of subgroups used in comparisons.

Next, the researchers plan to validate their ascertainment of side effects against the Adverse Event Profile questionnaire and conduct the same analysis in a larger, external set of patients at another tertiary referral center before testing it in the general epilepsy population.

None of the authors had relevant disclosures.