Visualizing Time-Varying Effect in Survival Analysis: 5 Complementary Plots to Kaplan-Meier Curve

BACKGROUND: Kaplan-Meier (KM) curve has been widely used in the field of oxidative medicine and cellular longevity. However, time-varying effect might be presented in KM curve and cannot be intuitively observed. Complementary plots might promote clear insights in time-varying effect from KM curve. METHODS: Three KM curves were identified from published randomized control trials: (a) curves diverged immediately; (b) intersected curves with statistical significance; and (c) intersected curves without statistical significance. We reconstructed individual patient data, and plotted 5 complementary plots (difference in survival probability and risk difference, difference in restricted mean survival time, landmark analyses, and hazard ratio over time), along with KM curve. RESULTS: Entanglement and intersection of two KM curves would make the 5 complementary plots to fluctuate over time intuitively. Absolute effects were presented in the 3 plots of difference in survival probability, risk, and restricted mean survival time. Changed P values from landmark analyses were used to inspect conditional treatment effect; the turning points could be identified for further landmark analysis. When proportional hazard assumption was not met, estimated hazard ratio from traditional Cox regression was not appropriate, and time-varying hazard ratios could be presented instead of an average and single value. CONCLUSIONS: The 5 complementary plots with KM curve give a broad and straightforward picture of potential time-varying effect. They will provide clear insight in treatment effect and assist clinicians to make decision comprehensively.