What Does Median Not Reached Mean in Clinical Trials?

“Median not reached” means that fewer than half of the patients in a group have had the event being measured.

In an overall survival analysis, the event is death. If median overall survival is not reached, fewer than 50% of patients in that group have died by the time of analysis.

In a progression-free survival analysis, the event is disease progression or death. If median PFS is not reached, fewer than 50% of patients have progressed or died.

Simple example

Imagine a cancer trial with 100 patients in one treatment arm. If 45 patients have died by the analysis date, median overall survival has not been reached. The trial can say “median OS not reached.” If 50 or more patients have died, the median can usually be estimated.

Why “median not reached” sounds bullish

In biotech press releases, “median not reached” often sounds very positive. It can suggest that patients are living longer than expected. Sometimes that is true. If the control arm has reached median survival and the experimental arm has not, that can be a strong sign of benefit.

But not always. “Median not reached” can simply mean that the data are immature. It can also happen if follow-up is short, many patients were enrolled recently, or the analysis occurred before enough events accumulated.

The most important distinction: arm-level vs pooled

Investors must distinguish between experimental arm median not reached, control arm median not reached and pooled median not reached.

In a blinded trial, the company may only know pooled data. If pooled median survival is not reached, it does not prove that the treatment arm is performing well. It only means that fewer than half of all patients combined have had the event.

Median not reached can still fail

A trial can have “median not reached” in the treatment arm and still fail if the control arm also performs well. For example:

Treatment arm mOS: not reached
Control arm mOS: not reached
Hazard ratio: 0.95
P-value: not significant

This would not be a strong efficacy result, even though both medians are not reached.

Why follow-up time matters

If median follow-up is short, “median not reached” is less informative. The trial may simply need more time. This is common in event-driven trials, where analysis depends on events rather than a fixed date.

What investors should ask

When a company says “median not reached,” ask:

Which median was not reached?
Was it treatment arm, control arm or pooled?
How many patients were in the analysis?
What was the median follow-up?
How many events had occurred?
Was the trial randomized?
Was the data blinded?
What was the hazard ratio?
What did the Kaplan-Meier curves show?
Was there heavy censoring?

Frequently asked questions

What does median not reached mean in a clinical trial?

It means fewer than half of the patients in a group have had the event being measured, such as death in an overall survival analysis, or progression or death in a progression-free survival analysis. Because fewer than 50% of patients have had the event, the survival curve has not crossed the halfway point, so no median can be calculated yet.

Is median not reached always a good sign?

Not always. It can suggest patients are living longer than expected, especially if the control arm has already reached its median and the experimental arm has not. But it can also simply mean the data are immature, follow-up is short, or too few events have accumulated. On its own it is not proof that a drug works.

Can a trial with median not reached still fail?

Yes. A trial can show median not reached in the treatment arm and still fail if the control arm also performs well. If both medians are not reached, the hazard ratio is near 1.0 and the p-value is not significant, the result is weak despite the headline. Median not reached is most meaningful when it is arm-level, mature and supported by a strong hazard ratio.

Bottom line

“Median not reached” can be encouraging, but it is not proof of success by itself. It is most meaningful when it is arm-level, mature, randomized and supported by a strong hazard ratio.