Clinical trials typically collect longitudinal data, data that is collected repeated over time, such as labs, scans, or patient-reported outcomes. Due to a variety of reasons, this data can be missing, whether a patient stops attending clinical visits (i.e., drop-out), or misses assessments intermittently. Understanding the reasons for missing data as well as predictors of missing data can aid in determination of the missing data mechanism. The analysis methods employed are dependent on the missing data mechanism and may make certain assumptions about the missing data itself. Methods for non-ignorable missing data, which assumes that the missing data depends on the missing data itself, make stronger assumptions and include pattern mixture-models and shared parameter models. Missing data that is ignorable after adjusting for other covariates, can be analyzed using methods that adjust for covariates, such as mixed effects models or multiple imputation. Missing data that is ignorable can be analyzed using standard approaches that require complete case data, such as change from baseline or proportion of patients who declined at a specified time point. In clinical trials, truly ignorable data is rare, resulting in additional analysis methods required for proper interpretation of the results. Conducting several analyses under different assumptions, called sensitivity analyses, can determine the extent of the impact of the missing data.

This content is only available as a PDF.

© The Author(s) 2021. Published by Oxford University Press on behalf of the Society for Neuro-Oncology and the European Association of Neuro-Oncology. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.