- © 2007 Canadian Medical Association or its licensors
We believe that the survival benefit of revascularization in patients with heart failure was probably overestimated in the study by Ross Tsuyuki and colleagues1 because of time-dependent bias.2 Time-dependent bias can occur whenever patients are assigned to treatment groups after the start of observation. Patients who experience an event early in the observation period will not receive the treatment being studied. As a result, the outcome risk in the untreated group is inflated and the relative benefit of treatment is exaggerated.
It is possible that this phenomenon occurred in the present study.
Patients were assigned to the revascularization group if they received treatment during the first year of observation. If patients had died at any time during the first year of observation before treatment, they would have been assigned to the control group. Also, if patients had experienced any other significant event (e.g., stroke) during the first year of observation, they would have been less likely to undergo revascularization. Either of these events would have made outcomes appear to be worse in the group of patients who did not undergo revascularization.
That the survival curves of the treatment groups appear to separate primarily during the initial year of observation suggests that a time-dependent bias might have played a strong role in the study's results. We strongly suggest that the analysis be repeated using time-dependent covariates to account for this potential bias.3