In a cohort analysis, participants are classified as 'treated' only if they survive long enough to receive treatment, and the time before treatment is counted in the treated group. What bias may result?