Abstracts

Rationale: The 6 Hz test is used to screen potential anticonvulsant compounds, but there is no standard statistical analysis for the 6 Hz test when it is applied to animals treated with dietary interventions. Traditionally, a fixed current is applied to mice after they are treated with anticonvulsant compounds administered at doses ranging over several orders of magnitude, yielding an ED50 for comparing anticonvulsants. However, the concept of ‘dosage’ is limited when evaluating dietary therapy as the anticonvulsant, since mice only tolerate a limited range of nutrient content. Therefore, we investigated the underlying statistical basis for comparing anticonvulsant diets in the 6 Hz test by evaluating different statistical parameters to establish accurate values for various outcome variables (e.g., in a power analysis). Here, we describe the statistical analyses of the 6 Hz test applied at a range of currents to compare a ketogenic diet (originally designed to mimic the beneficial effect of fasting on seizures) and a calorie-restricted diet of normal rodent chow. Methods: Male NIH Swiss mice (3-4 weeks old, post-weaning) were fasted overnight before consuming either a ketogenic diet ad-lib, normal rodent chow fed ad-lib, or normal chow with an alternating feeding/fasting schedule every other day for a total of 10-12 days. Diets were compared by probit analysis and different assumed probability distributions (e.g., normal, lognormal) using Minitab 15. Results: Mice consuming the ketogenic diet were protected against 6 Hz-induced seizures compared to mice eating normal rodent chow, as we reported previously. In contrast, intermittent fasting made mice more susceptible to 6 Hz-induced seizures. Residuals (i.e., difference between experimental values and those predicted from the probability distribution) were lowest overall for logarithmic distributions, suggesting they are most appropriate for inter-diet comparisons. The differences between slopes of the probability functions showed a trend toward being significantly different with the normal and logistic distributions (0.14 > P > 0.07), suggesting that other distributions (i.e., lognormal and loglogistic) may be better for comparing the diets over a wide range of currents. Conclusions: The ketogenic diet protected against 6 Hz seizures, while calorie restriction (intermittent fasting) increased susceptibility to seizures, regardless of the assumed underlying distribution used for the comparison. These data fit a lognormal or loglogistic distribution better than normal, logistic, or Weibull distributions and provide a rigorous estimate for designing future studies. We conclude that logarithmic distributions, commonly used in the analysis of anticonvulsant compounds, also have valid applications for the study of dietary therapies used to treat epilepsy. Supported by K12NS001696 (ALH) and NS037402 (JMH).