Attrition bias refers the the unequal loss of participants from different groups in a trial. Patients are lost to follow up in almost every trial. However, when different numbers of patients leave from each group, or when patients leave each group for different reasons, the trial groups may end the trial imbalanced, even if the trial started fairly.
As a general rule of thumb, we like to see trials with less than 5% of patients lost to follow up, and consider trials that lose more than 20% of patients to be at high risk of bias. Of course, those numbers have to be considered in the context of the trial.
Some studies assess the potential impact of attrition bias by performing a “worst case scenario” analysis. They assume that every patient who was lost to follow-up had the worst possible outcome. If the results of this analysis don’t change the study’s outcomes, then loss to follow-up is less likely to be a concern.
Attrition bias is one reasons that intention to treat analyses are so important. It is impossible to know the outcomes of every patient lost or why they left the trial, so it is important to analyze all patients in the groups to which the were randomized to limit the impact of attrition bias.
This post is part of a series of posts on bias in medical research. You can find the whole bias catalogue here.
Attrition bias is a type of selection bias.
Morgenstern, J. Attrition bias, First10EM, April 7, 2018. Available at: