design and not an inevitable result of the condition, treatment, or behavior. Unfortunately, few of the studies examined by the committee obtained outcome information after a patient stopped treatment or during post-treatment follow-up. Because a very high percentage of patients—typically 20 to 50 percent—typically dropped out of these studies, large fractions of outcome data were missing.
Over the past three decades, analytic approaches to handling missing data have matured, with multiple imputation and mixed-model repeated measurement (MMRM) and similar approaches being implemented in standard software and commonly used by biostatisticians in many fields (Little and Rubin, 2002; Molenberghs and Kenward, 2007). Unfortunately, the most common way missing data were handled in the literature reviewed was to use the last recorded outcome as the final outcome in a patient who dropped out—also known as the “last observation carried forward”