Figure 1.

Example of Kaplan-Meier curves (A) with missing visual elements (MVE) and (B) without MVE. Data were fabricated using a random number generator for a hypothetical randomized controlled trial comparing progression-free survival between two groups, assessed every 6 months following 1:1 randomization; tick marks represent censoring. In (A), the y-axis is restricted to only survival probabilities between 0.50 and 1.00, which exaggerates the visual differences between groups and misrepresents the overall outcomes of the patients. In (B), the number at risk interval shows that the experimental group has fewer patients than the control group following 1:1 randomization. This scenario may occur in per-protocol analyses when an experimental therapy has considerable upfront toxicity, resulting in a systematic loss of patients compared with intention-to-treat analysis. The number at risk table in (B), which includes the number censored, also suggests informative censoring in the control arm at 6 months compared with the experimental arm (14% vs 5%), which may add bias in favor of the experimental arm.