Loss of subjects during a clinical trial may lead to bias in the results.
The ideal is that everyone enrolled in a study completes it. A small number of subjects may not pose a problem. A problem can arise when certain conditions are present.
Sources of bias due to attrition:
(1) the total number lost is too high
(2) the number lost in one group exceeds the loss in another creating an imbalance
(3) the type of patients lost creates an imbalance and influences confounding
(4) subjects are lost due to unrecorded adverse effects or death associated with therapy
(5) subjects who drop out may have more severe disease, skewing the remaining subjects to milder disease with better response to the intervention
A general rule of thumb is that < 5% attrition is associated with low bias and > 20% can be significant. However, other factors may influence the impact.
There are ways of dealing with missing data but these all require some level of guessing.
