Theory 22/Clinical Trials Methodology
Attrition Bias Is a Ratio
The bias introduced by trial dropouts is set by the ratio of differential dropout to the event rate, so a small imbalance can badly distort a trial with few events and barely matter in one with many.
01The mechanism
Dropouts remove outcomes from the analysis. When events are rare, each removed event is a large share of the total, so a small differential has a large effect on the estimate.
The ratio determines how much the missing data can move the result, which is why the same dropout rate is tolerated in some trials and fatal in others.
02What this is not
Those set a fixed acceptable rate. This scales tolerance to the event rate, which explains why the same rate matters differently.
03Predictions
- P1Bias should track the ratio.
- P2Low-event trials should be more sensitive to small imbalances.
04Refutation
Bias tracking dropout rate alone would favour fixed thresholds.