How to Read a Cancer Trial and Tell If Patients Live Longer

Here is a headline you will see a version of most weeks. "New drug shrinks tumors in 40% of patients with advanced cancer." It sounds like great news. Before you believe it, ask one question. Did the patients actually live longer?

"The drug works" can mean a few different things. It can mean the tumor got smaller. It can mean the cancer took longer to grow back. Or it can mean people lived longer, and felt well while they did. Those are not the same claim, and a trial only proves the one it was built to measure. That one measure is the trial's primary endpoint, the single result it is powered to prove and judged on. Once you know how to read it, you can tell in about ten seconds how much a cancer result is really worth.

This brief is a plain guide to understanding the results of cancer clinical trials. It walks through the endpoints one at a time, then hands you a simple way to place any trial on a spectrum of confidence. Has it proven patients live longer, made it likely, or only made a promise?

The question underneath every cancer trial

What patients want is simple. Live longer, and feel well while doing it. So as you read a trial, keep asking one thing. Does this result actually tell me whether patients live longer?

The clinical trials that measure overall survival as their primary endpoint can answer that question directly. Overall survival is exactly what it sounds like. It is how long patients live, from the day they start treatment until the day they die, of any cause. If a trial improves overall survival, the argument is over. Patients lived longer.

The catch is that overall survival is slow. To measure it, you have to wait for the deaths, and that can take years. So trials often measure something faster instead, a stand-in they hope predicts survival. Progression-free survival, response rate, and pathologic complete response are the ones you will see most, with circulating tumor DNA now joining them. A drug can show improvement on one of these stand-ins but still not extend life meaningfully.

The endpoints, from strongest to fastest

Here is the whole family, tied together by one rule. The stronger the proof, the longer it takes to get. Trade speed for certainty and you land on the endpoint you are looking at.

As you go down the list, you get your answer faster and you can trust it less. And the counts are lopsided. The fastest, weakest endpoint (tumor response) is primary in far more trials than the definitive one (survival).

Those counts come from the TheraRadar oncology index, our curated extract of ClinicalTrials.gov. It holds 42,864 active and recent cancer trials, about 29% of the 148,000 we track, and cancer is the single largest field in the registry. Cancer trial starts in it rose 62% from 2020 to 2025.

A confidence ladder for cancer trial results

Take any cancer result and ask which endpoint it is built on. Make sure it is the primary endpoint, the one the trial was designed and powered to prove. A win on a secondary endpoint, or in one subgroup, is only a hint. That question tells you how much to believe the claim that patients live longer.

The hazard ratio, in plain terms

Cancer results almost always come with a number like HR 0.50. It looks technical, but it is simpler than it seems. The hazard ratio compares the risk in the two groups at any given moment. Here is how to read it.

One habit worth building is to always check which endpoint the hazard ratio belongs to. An HR of 0.50 on overall survival is a triumph, because it halves the risk of death. The same 0.50 on progression-free survival tells you the cancer was slower to grow, which is worth having but says nothing on its own about how long the patient lives.

Why a fast endpoint is only a promise

The FDA can approve a cancer drug early, on a fast endpoint like tumor shrinkage or delayed progression, before anyone knows whether patients live longer. It is called accelerated approval, and the deal is simple. The drug reaches patients sooner, and the company has to run a follow-up trial to prove the survival benefit is real. Sometimes that benefit turns out to be real, and sometimes it does not.

Often there is no comparison group at all. In a single-arm trial, everyone gets the drug and the result is judged against a preset bar, usually a response rate. These exist for good reason. When a cancer is rare, aggressive, and out of standard options, randomizing patients to a placebo or a weaker treatment can be impractical or hard to justify, so a clear response signal is allowed to stand on its own for accelerated approval, with a confirmatory trial owed later. The problem still remains. With no control arm, nothing is being measured against survival, so a single-arm result is only a promise.

Take Avastin (bevacizumab, from Genentech) in breast cancer. In 2008 the FDA gave it accelerated approval for metastatic breast cancer on a progression-free survival gain of about five and a half months. That is a real improvement in a fast endpoint. But the follow-up trials showed no gain in overall survival, and more side effects. In 2011 the FDA pulled the breast-cancer approval. The delay in progression had been real, but it never became longer life.

A fast endpoint is a bet that survival will follow, and you find out later whether the bet paid off. So when you see an approval built on tumor response, the honest reading is that the drug might help people live longer, and the survival data will settle it.

What the survival curve shows

Survival results usually get reported as a single number, the median. That is the point where half the patients have passed and half are still going. It is a fair summary, but it hides the shape. The full picture is a survival curve (you will see it called a Kaplan-Meier curve), which tracks the share of patients still alive over time.

A survival curve shows two things a median cannot.

• The tail. With some immunotherapies, a slice of patients keep going for years, far past the median. That long flat tail on the right is the real story, and a single median number erases it.
• Crossover. In many trials, patients in the control group are allowed to switch to the new drug once their cancer grows. That is the humane thing to do, but it can flatten the survival gap, because the control group ends up getting the drug too. So a weak-looking survival result is sometimes weak for a good reason.

What statistical significance leaves out

You will see the phrase statistically significant, usually tied to a p-value below 0.05. All it means is that the result is probably not a fluke. It says nothing about how big the effect is. A drug can be statistically significant and still add only two weeks of life, at a real cost in side effects and dollars. So once you know the effect is real, the next thing to ask is whether it is big enough to matter. For that, look at the hazard ratio and how many months were actually gained.

The confidence ladder in five real cancer trials

Five real trials, from a proven survival win to a promise that failed.

Five questions to ask of any cancer headline

1. Which endpoint? Overall survival, or a faster stand-in?
2. If it is a stand-in, has survival been shown yet, or is this still a promise?
3. What is the hazard ratio, and on which endpoint?
4. How big is the benefit in plain terms (months, or percentage points)?
5. Compared against what? A real standard treatment, or nothing?

Once you can answer those five, you understand the trial, whatever the headline happens to say.

Trial details and results are drawn from the public ClinicalTrials.gov and FDA records for each study. Trial counts come from the TheraRadar oncology index, a curated subset of ClinicalTrials.gov (2026-06). This is an educational reference and does not constitute medical advice.