This question investigates the likelihood that loss to follow-up was biased. To ensure a lack of bias, the study should include an intention-to-treat analysis. It also assesses whether the group assignment affected the likelihood of complete outcome reporting. Failure to conduct both intention-to-treat and per protocol analyses in studies with substantial outcome censoring would raise concern for attrition bias.
- Were outcome data available for all, or nearly all, participants randomized?
- Do the proportions of missing outcome data differ between intervention groups?
- Is it likely that missing data is associated with outcome (e.g., non-responders dropped out of trial)?