SOC is standard of care for a type of cancer drug, often some type of chemotherapy. FLIN is a new drug treatment (Made up name). OS is overall survival and it is the time that you either die or drop out of the study for other reasons, or the study period ends. RFS is relapse free survival and is the time to relapse of cancer, death, of discontinuation in the study due to dropping out or the study period ending.
This is survival analysis and it looks at a population of patients. All the patients starts out both alive and relapse free. When you make a plot, you start at 1.0 or 100 percent. As you go through time, people either relapse, die, or are censored, and the curve moves down and eventually it hits zero, meaning either everyone has relapsed or died.
If you look at two curves, the higher one is better than the lower one. The lower one means that more bad things (death, relapse) are happening. A cancer drug shows that it’s better by plotting higher.
The results of the survival estimates are used in cost effectiveness models. In most cases the data spans three to six years. However, the time horizon on the cost effectiveness is often 20 or more years, even though with some cancers and other diseases patients are unlikely to live that long. The curves do extend out of sample, but the parametric estimates are the best guess of what will happen in the future.
The models generally compute discounted life years or utility from living, along with discounted costs. You compute these measures for both drugs, and find the ratio of the differences which is a statistic called the incremental cost-effectiveness ratio (ICER).
This type of analysis is used by agencies around the world to justify the purchasing of a new treatments. It is basically saying how much does it cost for each additional quality adjusted life year. If it cost less than $10,000 keep someone alive for an additional year the drug is probably a good deal, if it cost over $10,000,000 it is probably too expensive. In countries with Nationalized healthcare, they have to make a judgement on if they will pay for a new drug or not. This analysis helps them make their decision.
In my analysis when I extrapolate my survival plots my OS and RFS cross so RFS is higher than OS. However, by definition of the measure, this can’t happen, and is only happening because I’m extrapolating out of sample. I’m turning to to Bayesian analysis to keep this from happening. In my case I did it first by forcing the RFS and OS curves to have the same cure rate, and then I forced the OS to to have a slightly higher rate by adding a “d” which I forced to be a small positive number (0.02 to 0.06)