The term “landmark” in this case usually refers to a landmark in time. As in, “after 12 months, how many patients are still alive?” This particular landmark analysis would be referred to as the “12-month overall survival rate.” These are often useful for a few reasons. First, if you have a relatively favorable treatment area when it typically takes years to progress, that’s good for patients, but bad for cancer clinical trials, since the longer they take, the more expensive they are to run. Instead, you can get a sense of how the trial is going by looking at landmark analyses.
More importantly, if you have a particularly bad form of cancer, median survival might be poor. But if you perform a landmark analysis, you might be able to identify a treatment that helps a small group of patients. The development of immunotherapy has been a good example of this phenomenon, where most patients would fail to respond, giving bad “median” progression-free survival, but a minority of patients would experience very long responses compared with, say, chemotherapy. Therefore, you might have similar median progression-free survival, but the survival at 12 months shows you that 20% of the patients were still alive with the treatment, while 0% were alive with the chemotherapy.