How to Read a Biotech Press Release Without Getting Fooled

You read the headline: “Drug X met its primary endpoint in Phase 3.” The stock jumps 30% before the market opens. By the time you read the actual data on page 3 of the press release, the company has started dumping shares into the rally. You bought the spin.

The solution: treat every biotech press release as a marketing document with real data buried inside it. Your job: find six numbers in 60 seconds: primary endpoint result, p-value, confidence interval, control arm comparison, sample size, and safety signals. If the company is highlighting anything else, it is usually hiding a miss.

Step 1: Find the primary endpoint first, always

The primary endpoint is the outcome the trial was statistically powered to detect. It was pre-specified in the trial registration on ClinicalTrials.gov before the first patient was enrolled. If the trial met its primary endpoint, the press release will say so in the first sentence. If it says the trial “met key secondary endpoints” or “showed numerical improvement” and is quiet about the primary, the primary missed. That is not a positive trial.

When a company headlines a secondary endpoint, the primary likely failed. Secondary endpoints are exploratory, the trial was not powered to detect them, and a positive secondary in a trial that missed its primary is statistically suspect. The FDA almost never approves a drug on a missed primary with a positive secondary.

Real example: GSK’s camlipixant (CALM-2) for chronic cough missed its primary endpoint. The press release led with “numerical improvement” and patient-reported outcomes. See our GSK camlipixant CALM-2 failure writeup.

Step 2: Read the p-value AND the effect size

The p-value tells you whether the result could have happened by chance. The threshold is p < 0.05, below that, the result is statistically significant. But statistical significance is not the same as clinical meaningfulness. A p-value of 0.04 on a 2-point improvement on a 50-point scale is statistically significant and clinically meaningless.

Always read the effect size, how big the difference actually is, alongside the p-value. A 50% reduction in risk (hazard ratio 0.50, p < 0.001) is a meaningful effect. A 5% reduction in risk (hazard ratio 0.95, p = 0.04) is a marginal effect that may not survive a larger trial. The FDA cares about magnitude, not just significance.

Step 3: Read the confidence interval, not just the point estimate

The confidence interval (CI) tells you the range of plausible true effects. A 95% CI of 0.35 to 0.65 for a hazard ratio means we’re 95% confident the true effect is between a 35% and 65% risk reduction. A narrow CI means the result is stable. A wide CI means the result is noisy.

Red flag: If the lower bound of a 95% CI crosses 1.0 (for a hazard ratio or odds ratio) or 0 (for a difference measure), the result is not statistically significant even if the point estimate looks good. Companies sometimes headline the point estimate and bury the CI. If the CI crosses the null, the trial did not meet its endpoint by the standard statistical criterion, no matter what the press release says.

Step 4: Apply the Phase 2-to-Phase 3 translation problem

This is the single most expensive lesson in biotech investing. Phase 2 trials are smaller, often single-arm, and frequently use surrogate endpoints. Phase 3 trials are larger, randomized, controlled, and use clinical endpoints. Only about one-third of Phase 2 successes replicate in Phase 3 (BIO industry analysis 2011-2020; Phase II-to-III transition rate of roughly 30-45% across therapeutic areas, neurology as low as 27%).

A positive Phase 2 press release is not an approval signal. It is a hypothesis to be tested in Phase 3. Investors who buy Phase 2 readouts at full premium lose money about two-thirds of the time. See our biotech investing guide for phase pass-rate data.

Real example: Celldex’s barzolvolimab Phase 2 in chronic spontaneous urticaria showed biological activity (mast cell depletion). The clinical benefit, symptom improvement, did not follow. See our Celldex barzolvolimab PN failure writeup. Biological activity is not clinical benefit.

Step 5: Spot subgroup analysis traps

Subgroup analysis slices the trial population by characteristics (age, disease severity, biomarker status). It is useful for hypothesis generation, but it is also the most common way to spin a failed trial. When the primary endpoint misses in the overall population, companies often highlight a positive subgroup (“patients with high baseline disease severity showed a 40% response”).

Red flag: Post-hoc subgroup analyses, analyses done after seeing the data, have inflated false-positive rates and are not statistically valid. Always check whether the subgroup was pre-specified in the trial registration on ClinicalTrials.gov. If it wasn’t, the result is hypothesis-generating, not confirmatory. The FDA rarely approves on post-hoc subgroup analysis of a failed trial.

Real example: Ipsen’s Bylvay (odevixibat) BOLD Phase 3 in biliary atresia missed its primary endpoint. The press release highlighted “descriptive trends” in subgroups. See our Ipsen Bylvay BOLD Phase 3 miss writeup. “Descriptive trends” is press-release code for “the trial failed and we’re looking for anything positive to say.”

Step 6: Run the 60-second checklist

Before you trade on any clinical trial press release, answer these six questions:

Was the primary endpoint met? If the press release doesn’t say “met primary endpoint” in the first sentence, it didn’t.
Was the trial randomized and blinded? Open-label and single-arm trials are weaker evidence.
What was the control arm? Placebo, standard of care, or historical control? Historical controls are weaker. If the drug group had a 30% response and placebo had 25%, the drug isn’t doing much.
Sample size? A trial of 30 patients can show anything. A trial of 3,000 is harder to fool.
Effect size? Statistically significant does not mean clinically meaningful. Read the absolute difference, not just the p-value.
Safety signals? Grade 3/4 adverse events, discontinuations, deaths, were there more in the drug arm than the control arm? “Safety was consistent with the known profile” without numbers is a red flag.

Step 7: Verify against ClinicalTrials.gov

Every registered trial has a public record on ClinicalTrials.gov with the pre-specified primary and secondary endpoints, sample size, and trial design. Search the NCT number (usually in the press release or on the company’s IR page). Compare what the trial was designed to measure to what the press release headlines. If they don’t match, the company is spinning. This is the single highest-value habit in biotech investing. It takes two minutes, it is free, and it catches more spin than any other step. If a company changed endpoints between registration and readout, or downgraded the primary to secondary mid-trial, that is a major red flag.

Step 8: Understand the surrogate endpoint problem

A surrogate endpoint is a biomarker, tumor shrinkage, dystrophin production, or hemoglobin levels, used as a stand-in for a clinical outcome like survival or walking ability. Surrogates are faster to measure, which is why companies love them. They are also less reliable: a drug can improve a biomarker without improving how patients feel or how long they live.

The FDA may grant accelerated approval on a surrogate endpoint, but the company must run a confirmatory trial proving clinical benefit. Under FDORA (2022), the FDA now has expedited withdrawal authority if that confirmatory trial fails. See our accelerated approval guide for the mechanism. When you read a press release, check whether the endpoint was a surrogate or a clinical outcome. A positive surrogate result is a faster path to approval, but a riskier one.

Common mistakes

Buying the headline without reading the data. The data is on page 3. If you don’t read it, you’re trading the company’s narrative, not the science.
Confusing statistical significance with clinical meaningfulness. A p-value under 0.05 on a tiny effect is still a tiny effect.
Trusting a positive subgroup in a failed trial. Post-hoc subgroups are hypothesis-generating. The FDA almost never approves on them.
Treating Phase 2 like Phase 3. Only about one-third of Phase 2 successes replicate in Phase 3.
Not checking ClinicalTrials.gov. If the press release’s headline endpoint doesn’t match the pre-specified primary endpoint in the trial registration, the company is spinning.
Ignoring safety signals. A drug that works but kills 3% of patients will not get approved.

Final checklist

Primary endpoint result found in the first sentence (or its absence noted)
p-value AND effect size read together
95% confidence interval checked - does it cross the null?
Trial design confirmed: randomized, blinded, controlled
Sample size noted - under 100 is hypothesis-generating, over 1,000 is confirmatory
Safety data read - Grade 3/4 AEs, discontinuations, deaths
NCT number verified on ClinicalTrials.gov
Endpoint classified as surrogate or clinical
Position sized for a 50% loss on catalyst trades. See our FDA catalyst trading guide.