Clinical trial results often have holes, the result of volunteers missing appointments or dropping out. Enough missing data reduces the benefit of randomization and can skew treatment group comparisons. But trial design and analysis methods can be tweaked to better account for the shortfall.
Two elements are critical, according to a panel created for the FDA to help develop guidance on trial design and follow-up to minimize missing data and its effect on study results. The first is trial design and execution that limit the amount and impact of missing data. The second is scrutiny of investigator assumptions about the impact of missing data on estimates of treatment effects.
The panel developed recommendations, some of which appeared last month in Laboratory Equipment. It suggested that sponsors should specify statistical methods for handling missing data in their study protocols and make sure that clinicians understanding them. The panel believes that in nearly all cases, there are better alternatives to last observation or baseline observation carried forward imputation. In addition, the panel advocates that reporting requirements carry a mandatory component of sponsor findings on sensitivity to the assumptions about the missing data mechanism.
The panel also determined the FDA has a very large database that has not been tapped for its potential to inform the best practices for clinical trials.
- here's the article