|
Abstract:
|
The FDA Guidance on Migraine: Developing Drugs for Acute Treatment (draft guidance, 2014) has the following recommendations on safety exposure. To be counted in the long-term safety database, adult patients should treat, on average, a minimum of two migraine attacks per month, and the development program should include a database of chronic intermittent use in at least 300 patients using the drug for 6 months, and 100 patients using the drug for 1 year. We propose a multiple imputation method based on a nonhomogeneous Poisson process with random effects to predict whether the above exposure requirements can be met by any given date using data from an ongoing long-term safety study. This method allows the rate of migraine attacks to vary over time and across patients after adjusting for covariates. We propose to impute the time of migraine attacks between now and a prespecified future date for ongoing patients to guarantee the monotonicity of imputed number of attacks over time at the patient level.
|
