Reading a hazard ratio

The hazard ratio comes from a model that compares event rates across the entire follow-up period, not just at one time point; the confidence interval around it must exclude 1.0 for the difference to be considered statistically significant, and the closer to zero, the larger the effect. Because it is relative, the same hazard ratio can mean very different absolute gains: 0.70 might add one month to a median of three, or two years to a median of six, so it should always be read alongside the median difference and the survival curves. A hazard ratio also assumes the effect is roughly constant over time, which immunotherapy, with its delayed benefit and long tail, often violates.