Loading
The time by which half the patients in a group have died (or, for other endpoints, progressed). It is the midpoint of a spread, not a prediction for anyone: half live longer, some much longer.
Medians are used rather than averages because some patients live far longer than the rest and the average would be dragged up and undefined until the last patient dies; the median can be read off the survival curve at the point where it crosses 50%. Comparing medians between arms gives an absolute benefit in months, complementing the hazard ratio's relative measure, but it captures only one point on the curve: two treatments with the same median can differ hugely in the proportion alive at three years, which is where immunotherapy's long tail shows up. 'Median not reached' means fewer than half the patients had the event at the time of analysis, a good sign that is also frustratingly uninformative.
Shares Absolute versus relative benefit (number needed to treat), Quality-adjusted survival (QALYs and Q-TWiST), Kaplan-Meier curve, censoring and proportional hazards, Hazard ratio (HR).
Shares Reading a hazard ratio, Confidence interval, Quality-adjusted survival (QALYs and Q-TWiST), Hazard ratio (HR).
Shares Kaplan-Meier curve, censoring and proportional hazards, Progression-free survival (PFS), Overall survival (OS).
Shares Reading a hazard ratio, Confidence interval, Hazard ratio (HR).
Shares Hazard ratio (HR), Progression-free survival (PFS), Overall survival (OS).
Shares Hazard ratio (HR), Progression-free survival (PFS), Overall survival (OS).
Shares Kaplan-Meier curve, censoring and proportional hazards, Hazard ratio (HR), Overall survival (OS).
Shares Hazard ratio (HR), Progression-free survival (PFS), Overall survival (OS).