Clinical Trial Endpoints: Primary, Secondary, Surrogate

Problem

You read a biotech press release that says “the drug met its primary endpoint.” The stock jumps 30%. You buy. Three months later the FDA rejects the drug and the stock drops 60%. What happened? The endpoint the company “met” was a surrogate marker that didn’t predict real clinical benefit, and you didn’t know the difference.

Solution

Learn the four endpoint types: primary, secondary, surrogate, and composite. Treat the primary endpoint result as the only result that actually matters for approval.

Step-by-step

Step 1: Find the primary endpoint

The primary endpoint is the main outcome the trial was designed to measure. It’s pre-specified: the company committed to it in the trial registration on ClinicalTrials.gov before the first patient was dosed. If the primary endpoint fails, the trial fails, full stop. Secondary endpoint results cannot rescue a failed primary.

Real example: In 2025, GSK announced that camlipixant missed the primary endpoint in the CALM-2 Phase 3 trial for chronic cough. The press release led with “numerical improvement” and patient-reported outcomes. The stock dropped on the news because the primary endpoint, cough frequency reduction, was not met. The rest was spin. We covered that miss in our GSK camlipixant analysis.

Step 2: Read the secondary endpoints, but don’t be fooled

Secondary endpoints are additional outcomes that support the primary. They can strengthen the case for approval, but they cannot override a failed primary. When a press release highlights secondary endpoints while burying the primary, treat it as a red flag.

Common pattern: “The drug did not meet the primary endpoint of overall survival, but showed a statistically significant improvement in progression-free survival.” PFS is real data, but it’s secondary. The FDA will scrutinize the benefit-risk profile, and approval is not guaranteed.

Step 3: Identify whether the endpoint is a surrogate

A surrogate endpoint is a biomarker or intermediate measure that predicts clinical benefit but is not itself the clinical outcome. Tumor shrinkage (response rate) is a surrogate for survival. LDL cholesterol reduction is a surrogate for fewer heart attacks. Dystrophin production is a surrogate for motor function in Duchenne muscular dystrophy.

The FDA grants accelerated approval based on surrogates for serious conditions with unmet need, but the company must run a confirmatory trial to prove real clinical benefit. If the confirmatory trial fails, the FDA can withdraw the approval. Sarepta’s Elevidys (SRPT) was approved on a dystrophin-surrogate basis; the functional-benefit question nearly derailed it, and the FDA has since tightened its stance on DMD surrogates.

Surrogates are the most dangerous endpoint type for investors. They let a drug reach the market before anyone knows if it actually helps patients.

Step 4: Recognize composite endpoints

A composite endpoint combines multiple outcomes into a single measure. The classic example is MACE (major adverse cardiovascular events), which counts cardiovascular death, non-fatal heart attack, and non-fatal stroke as one endpoint. Composite endpoints increase statistical power, but they can hide which component is doing the work. Some drugs reduce MACE overall while showing no real effect on mortality, driven entirely by fewer non-fatal events.

When you see a composite endpoint, ask: which component is driving the result, and is that component clinically meaningful?

Step 5: Separate statistically significant from clinically meaningful

A trial can show a statistically significant p<0.001 improvement on a 50-point symptom scale that turns out to be a 2-point difference. Patients can’t feel a 2-point difference. That result is statistically significant and clinically meaningless.

Look at the effect size, the actual difference between the drug arm and the control arm, not just the p-value. A 15-point improvement on the same scale is clinically meaningful. A 2-point improvement is not, no matter how small the p-value.

Arrowhead’s plozasiran Phase 3 (SHASTA-3/4) reported a median triglyceride reduction of 57-63% versus baseline, a large, clinically meaningful effect on the primary endpoint. That’s the kind of number that actually matters. We broke down that readout in our Arrowhead plozasiran analysis.

Step 6: Learn the endpoint hierarchy

The hierarchy matters when results conflict:

• Primary fails → trial fails. Walk away.
• Primary meets, secondary fails → FDA scrutinizes benefit-risk. Approval possible but uncertain.
• Surrogate met → accelerated approval, confirmatory trial required. The stock pops, but the risk isn’t gone.
• Composite met → check which component drove the result before trusting the headline.

Step 7: Read the press release the right way

For every endpoint the press release mentions, find four things: (1) the exact endpoint name, (2) the treatment effect versus control, (3) the p-value and 95% confidence interval, and (4) whether the endpoint was pre-specified or post-hoc. Post-hoc endpoints (invented after the data was unblinded) are marketing, not science.

This is exactly the skill we teach in our companion guide on how to read a clinical trial press release. Endpoints are the vocabulary; that guide is the field manual.

Common mistakes

• Buying on “met primary endpoint” without checking if the endpoint was a surrogate. The stock pops on the headline, then drops 40% when the confirmatory trial fails. Check the endpoint type before buying.
• Treating a secondary endpoint win as a primary win. A drug that missed its primary but “showed benefit” on a secondary is a drug with a failed trial. The FDA does not have to accept secondaries.
• Ignoring the confidence interval. A point estimate that looks great but has a 95% CI crossing 1.0 (hazard ratio) is not statistically significant. Companies highlight the point estimate and bury the CI.
• Trusting post-hoc subgroup analysis. If the overall trial failed but a subgroup “showed benefit,” the result is underpowered and usually a false positive. The FDA rarely approves on subgroup analysis of a failed trial.
• Forgetting that endpoint data can be questioned post-approval. Amgen’s TAVNEOS faced an FDA data-manipulation review years after approval when the integrity of the endpoint data came into question. We covered that here. Approval is not the end of endpoint risk.

Final checklist

• Is the primary endpoint pre-specified on ClinicalTrials.gov?
• Did the trial meet the primary endpoint, not just secondaries?
• Is the primary endpoint a surrogate or a clinical outcome?
• Is the effect size clinically meaningful, not just statistically significant?
• Does the 95% confidence interval exclude the null (1.0 for HR, 0 for differences)?
• Is the result from a pre-specified analysis, not a post-hoc subgroup?
• If surrogate-based, is a confirmatory trial underway?

Endpoints are the one thing in biotech investing you cannot skip. The PDUFA date and the trial readout only matter if you know what was actually measured. Start with our foundational investing guide if you’re new, then come back here before every catalyst trade.