Work out dropout rate impact instantly with clear inputs, formula shown and shareable results.
Attrition erodes power. Losing 54 of 300 participants is an 18% dropout rate, leaving 246 for analysis and cutting power from the planned 90% to about 83%. Beyond roughly 20% loss, CONSORT guidance warns that bias from differential dropout becomes a bigger concern than the loss of power itself.
Dropout rate
rate = dropouts / randomised
Retained power
recompute power with the analysed sample against the originally targeted effect
Under 5% is unlikely to bias results and over 20% is a serious threat. Between the two, sensitivity analyses are needed.
Because if one arm loses more participants, or loses different kinds of participant, the comparison itself becomes biased - which no sample size can fix.