Why Overall Survival Is the Gold Standard in Cancer Trials

Overall survival is the gold standard in cancer trials because it measures the one outcome that matters without ambiguity: whether a treatment helps people live longer. It counts every death from any cause, it needs no interpretation by a radiologist or a review committee, and it cannot be inflated by how you define a scan or when you schedule it. Response rate and progression-free survival are surrogates. They read out faster and can hint that a drug is active, but they frequently fail to predict whether patients actually live longer, which is why the U.S. Food and Drug Administration and independent researchers treat overall survival as the most reliable endpoint in oncology.

This article is educational and is not medical advice.

What the three endpoints actually measure

Three endpoints dominate cancer trials, and they answer different questions.

Response rate (often the objective response rate, or ORR) is the fraction of patients whose tumors shrink by a defined amount on imaging. It is a measure of tumor activity at a snapshot in time. A drug can shrink tumors impressively and still not change how long anyone lives.

Progression-free survival (PFS) is the time from randomization until the tumor grows past a threshold or the patient dies, whichever comes first. It captures disease control over time, but the "progression" half depends on scan timing, imaging criteria, and who is reading the images.

Overall survival (OS) is the time from randomization until death from any cause. There is no adjudication and no judgment call. A patient is alive or is not.

The FDA guidance titled Clinical Trial Endpoints for the Approval of Cancer Drugs and Biologics, most recently revised in December 2018 as a replacement for the 2007 version, lays out this hierarchy. It describes overall survival as the most reliable cancer endpoint and notes that it is generally the preferred basis for demonstrating clinical benefit, while response rate and progression-free survival are surrogate endpoints that may support approval under certain conditions.

Why overall survival resists manipulation

Two properties make overall survival hard to distort.

First, the event is death, which is unambiguous and needs no expert interpretation. Surrogate endpoints depend on measurement rules. If scans are done every six weeks in one arm and every twelve in another, measured progression can differ for reasons that have nothing to do with biology. This is called assessment bias, and it inflates PFS without helping a single patient.

Second, overall survival captures the full balance of benefit and harm. A therapy might delay tumor growth yet carry toxicity that shortens life through a different mechanism. Surrogates that look only at the tumor miss that trade-off. Overall survival, because it counts death from any cause, does not.

The 2025 draft FDA guidance, Approaches to Assessment of Overall Survival in Oncology Clinical Trials, published in the Federal Register in August 2025 and open for public comment through October 2025, reinforces this point from the safety side. As a draft, it is a proposal and not binding, but it recommends that sponsors collect and prespecify overall survival data even in trials designed around a surrogate, so that a hidden survival harm cannot slip past unnoticed. The logic is straightforward: if a drug is quietly costing lives, only an endpoint that counts deaths will reveal it.

When surrogates mislead

The gap between surrogate and survival is not hypothetical. The peer-reviewed literature documents trials where response rate and progression-free survival pointed one way and overall survival pointed another.

A 2020 review in the World Journal of Clinical Oncology by Aykan and Ozatli, which explicitly calls overall survival the gold standard, walks through concrete examples. In the FIRE-3 trial in metastatic colorectal cancer, objective response rate did not differ significantly between the two treatment arms and progression-free survival was similar, yet overall survival was significantly better with one regimen. Neither surrogate predicted the survival difference that mattered. In another trial the authors cite, progression-free survival was significantly superior with an added agent while response rates were essentially the same, showing how loosely these measures can track each other.

The broader pattern is sobering. Analyses of drugs approved on surrogate endpoints have repeatedly found that many show no overall survival benefit when survival data finally mature. A surrogate is only trustworthy when it has been formally validated against survival in that specific disease and setting, and validation in one cancer does not transfer automatically to another.

Why surrogates still have a role

None of this means surrogate endpoints are worthless. They are faster, which matters when patients with aggressive disease cannot wait years for survival data. That is the rationale behind accelerated approval, where a drug can reach patients based on a surrogate reasonably likely to predict benefit, with the expectation that confirmatory evidence follows. The FDA framework keeps overall survival as the anchor: surrogates can open the door, but converting to full approval generally leans on confirming that the benefit is real and durable.

The honest position is that surrogates are a reasonable bet under uncertainty, not proof. When you read that a drug improved response rate or progression-free survival, the useful question is whether that surrogate has been validated as a survival predictor in this disease, and whether the survival data are simply immature or actually flat.

How to read an oncology result

A practical checklist follows from all this. Identify which endpoint the headline rests on. If it is response rate or progression-free survival, ask whether overall survival was measured, what it showed, and whether it was even mature enough to interpret. Watch for imbalanced scan schedules or open-label designs that can bias a surrogate. And treat a large surrogate effect with a flat or unreported survival curve as a reason for caution, not celebration. The endpoint that survives all of these questions is overall survival, which is precisely why it remains the standard against which the others are judged.