Tricision Patient guide
Patient guide

How to read glioblastoma treatment numbers: survival, progression and what a "median" means

A "median survival of 15 months" means half the patients in that group lived longer than 15 months and half did not; it is not a prediction for you. Progression-free survival counts time until the tumor grows on MRI, overall survival counts time until death, and the two cannot be compared with each other. Numbers from different studies can only be compared cautiously, because the patients, the era and the way they were selected differ. Ask for the patient numbers, the follow-up length and the comparison group behind every figure.

Medically reviewed by Prof. Youwen He, MD, PhD · Last updated 16 October 2026

Median is not average and not a prediction

Survival in glioblastoma is reported as a median because the distribution is skewed: many patients cluster around the middle, and a smaller number live for years. The median is the point where half the group has had the event. It says nothing about where any individual will fall. Two groups with the same median can have very different numbers of long survivors, which is why the two-year and five-year survival rates, and the shape of the Kaplan-Meier curve, carry information the median does not.

What a "median" looks like on a survival curve
0%25%50%75%100%Median: half the group has had the event by hereLong tail: a few live for years; the median says nothing about them01224364860Months from the starting pointPatients still alive or recurrence-free

Illustrative, not patient data

Overall survival versus progression-free survival

Overall survival (OS) is time from a defined starting point (diagnosis, surgery, randomization or first dose) until death from any cause. It is the hardest endpoint to argue with but takes the longest to measure. Progression-free survival (PFS) is time until the tumor is seen to grow on MRI or the patient dies. It is available sooner and is less affected by what happens after progression, but it depends on how often scans are done and on the criteria used to call growth, and it can be confused by pseudoprogression. A treatment can lengthen PFS without lengthening OS, as bevacizumab does in glioblastoma. Never compare one treatment's PFS with another's OS; they measure different things.

Where the starting point is set matters

Time counted from diagnosis will always be longer than time counted from the first vaccine dose, because the vaccine is given months later. When a program reports time to recurrence from its first dose, as we do, it cannot be lined up against a trial that counts from surgery. Ask what the clock starts on.

Two clocks that are often confused
Diagnosis / surgeryRadiotherapyFirst vaccine doseRecurrence on MRITrial counting from surgery: 12 monthsProgram counting from first dose: 8.5 months

Same patient, same event, two different numbers. Ask where the clock starts.

Who was in the group

Patients who enter trials or travel abroad for treatment are younger, fitter and more often completely resected than the average patient. A group selected this way will outlive an unselected population even with identical treatment. This is why single-arm results, including early results from any vaccine program, cannot prove benefit on their own, and why a randomized comparison within the same study is the standard of proof. "External" or "historical" controls, as used in the DCVax-L Phase 3 trial, are better than nothing but remain debated.

Follow-up length and data cuts

A median cannot be calculated until half the group has had the event. Early in a study the median is "not reached", which sounds good but means the data are immature. A figure reported "at the current data cut" will change as follow-up lengthens; in a small group it can change a lot. Our own Phase I figure of 15.4 months is of this kind and is labelled as such.

Six questions to ask about any number

  1. What exactly is being measured: OS, PFS, time to recurrence, response rate?
  2. From what starting point?
  3. How many patients, and how were they selected?
  4. What is the comparison group, and was it concurrent and randomized?
  5. How long is the follow-up, and is the median reached?
  6. Is the result published and peer-reviewed, company-reported, or from an intermediary?

A provider who can answer all six is giving you data. One who cannot is giving you marketing. We publish our answers on our evidence section and on the science page.

Questions families ask

What does "hazard ratio 0.7" mean?

That at any moment, patients on the new treatment had about 70 percent of the risk of the event compared with the control group, a 30 percent relative reduction. It summarizes the whole curve rather than one point, and it only makes sense in a randomized comparison.

Why do older studies show shorter survival than newer ones?

Better imaging, surgery, supportive care and the removal of IDH-mutant tumors from the glioblastoma category under WHO 2021 all push modern figures up. Comparing a 2026 series with a 2005 benchmark flatters the newer series; this is why we state the limitation on every page.

Is a 9-year survivor proof that a treatment works?

No. Long survivors exist with standard treatment alone, especially with IDH-mutant tumors. A single case shows what is possible, not what is typical. Our published 9-year case is IDH1-mutant and we say so.

Related guides

Sources

  1. Liau LM, et al. DCVax-L Phase 3 externally controlled cohort trial. JAMA Oncology 2023 doi.org
  2. Wen PY, et al. RANO criteria for response assessment in high-grade gliomas. J Clin Oncol 2010 doi.org
  3. Stupp R, et al. NEJM 2005 doi.org

Considering treatment in China?

Send the patient's records for a remote review. A coordinator confirms receipt within one business day; the multidisciplinary team's report and a written estimate follow in about a week, before any decision or payment.

Request a records reviewAbout the program