The 10 clinical trials randomized 3078 subjects with a total of 3719 person-years of follow-up (). Consistent with the inclusion criteria, patients were at a mild to moderate stage of AD at randomization (mean [SD] MMSE score, 20.2 [3.9]; mean [SD] duration of dementia, 4.3 [2.6] years). Seventy-two percent were taking AchE-I overall. Of the 2238 individuals tested, 64% carried the APOE ε4 allele.
Demographic and Clinical Characteristics of Patients
Eighteen seizures were identified from the verbatim adverse event listings in the study: Cases identified were reported as seizure (n=13), possible seizure (n=2), probable seizure (n=2), or prodrome preseizure symptoms (n=1, included as a possible aura [partial seizure]). Eight of the 18 seizure adverse events were reported as serious adverse events. Four subjects discontinued study medication within 1 week of the seizure adverse event.
The incidence rate during 3719 person-years of follow-up was 484 per 100 000 person-years (95% CI, 287–764). Seizure rates within selected patient subgroups are shown in , although precision of subgroup estimates is limited by the low numbers of events. Subjects with greater cognitive impairment at baseline had a higher incidence rate: Subjects with an MMSE score less than 18 had an incidence of 1110 per 100 000 person-years, while those with an MMSE score of 18 or greater had an incidence rate of 257 per 100 000 person-years. Seventy-two percent of all subjects were taking AchE-I, and 89% of subjects with seizures were taking AchE-I at baseline; the exact confidence intervals for the seizure rate in subjects taking and not taking AchE-I broadly overlapped. The valproate study had the most reported seizures, likely owing to an overrepresentation of subjects with moderate AD as this study had a lower MMSE inclusion range (MMSE score range, 12–20) than the other studies.
By univariable Cox proportional hazards regression, significant univariable predictors of seizure were younger age (HR, 0.80; 95% CI, 0.70–0.91 per every 5 years of age), younger age at dementia onset (HR, 0.79; 95% CI, 0.70–0.90 per every 5 years of age), lower MMSE score (HR, 3.9; 95% CI, 1.5–10.1 for MMSE scores <18 compared with MMSE scores ≥18), memantine use (HR, 3.87; 95% CI, 1.12–13.49), and antipsychotic use (HR, 4.0; 95% CI, 1.5–10.3) (). Duration of dementia, APOE ε4 allele, treatment arm (omnibus test for placebo, active, or blinded), and cholinesterase inhibitor use were not significant univariable predictors of seizure. In a multivariable model including the factors age, MMSE score (continuous or <18 compared with an MMSE score ≥18), antipsychotic drug use, and memantine use, all factors except memantine use remained significant independent predictors of seizure occurrence, with attenuation of the effect estimates for MMSE score, antipsychotic use, and memantine use relative to the univariable analysis (). Age and antipsychotic use remained significant in the fully adjusted model; MMSE score, though not significant (P=.07), had an HR of 3.15 (95% CI, 0.93–10.75), with wider confidence intervals relative to the other models, suggesting collinearity with variables in the overparameterized model ().
Univariable and Multivariable Models of Seizures in Alzheimer Disease