Phase 3 Readouts: 8 Steps Before You Trade

A biotech you own just announced “positive Phase 3 results.” The stock is up 15% in premarket. Your phone is buzzing. Do you hold, add, or sell before the open?

The answer depends on what the data actually says, and most retail investors never get past the headline. Here is the eight-step framework for reading a Phase 3 readout in the 30 minutes before the market opens.

The one-sentence answer

Read the primary endpoint, the effect size, and the safety table before you read the CEO quote. The headline tells you how the market will react; the data tells you whether the reaction is right.

Step 1: Read the headline, then ignore it

“Met primary endpoint” is the only phrase in the headline that matters. Everything else is marketing. If the headline says “positive topline results” but never says “met primary endpoint,” that is a red flag. Companies that hit their primary endpoint say so, loudly, in the first sentence.

If the headline leads with “met key secondary endpoints” and is quiet about the primary, the primary missed. This is the single most common form of trial-data spin in biotech. For a deeper breakdown of press-release framing tactics, see our clinical trial press release reading guide.

Step 2: Identify the primary endpoint and whether it passed

The primary endpoint is the one metric the trial was statistically powered to measure. It is locked in at trial registration on ClinicalTrials.gov before the first patient is enrolled. The FDA bases approval on the primary endpoint. Secondary endpoints support the filing but do not replace a missed primary.

Real example: Summit Therapeutics (SMMT) ran the Phase 3 HARMONi trial for ivonescimab in EGFR-mutant lung cancer. The trial had strong progression-free survival data, but the primary overall survival endpoint missed in the initial analysis. Summit submitted the BLA anyway, framing the miss as technical. The FDA has historically been hostile to applications where the primary endpoint failed. That is why we called the SMMT PDUFA a coin flip: the missed primary is the central regulatory risk.

If the primary endpoint passed, move to Step 3. If it missed, the stock pop is a selling opportunity unless you have a specific thesis on why the FDA will accept secondary data.

Step 3: Check the p-value and confidence interval

The p-value tells you whether the result could be chance. The threshold is p < 0.05. Below that, the result is “statistically significant.” But two things matter more than the threshold itself.

First, how close is it? A p-value of 0.0412 barely cleared the bar. That is exactly what happened with Ionis (IONS) zilganersen in Alexander disease: the primary endpoint hit p=0.0412 in a 54-patient trial. The drug is likely to be approved because of Breakthrough Therapy designation and biomarker confirmation, but the marginal p-value is the core risk. See our zilganersen platform analysis for the full breakdown.

Second, read the confidence interval (CI). The CI tells you the range of plausible results. A hazard ratio of 0.76 with a 95% CI of 0.58 to 0.99 is statistically significant, but the upper bound barely excludes 1.0. If the CI crosses 1.0 for a hazard ratio or 0 for a difference, the result is not significant even if the point estimate looks good. Companies highlight the point estimate and bury the CI. Do not let them.

Step 4: Compare effect size to clinical significance

Statistical significance is not clinical significance. A p < 0.001 result with a 0.2 m/s walking-speed difference is statistically significant but clinically marginal. The FDA cares about whether patients feel better, live longer, or function better, not whether a p-value cleared a threshold.

BridgeBio’s (BBIO) BBP-418 FORTIFY trial in LGMD2I/R9 muscular dystrophy is a good example of both sides. The primary biomarker endpoint hit p < 0.0001, and the functional endpoint (100-meter timed test) showed a 0.27 m/s improvement versus placebo (p < 0.0001). That is both statistically and clinically meaningful in a population losing ambulation. Our BBP-418 PDUFA analysis walks through why the effect size supports approval. Compare that to a hypothetical drug that shows a 0.05 m/s improvement at p < 0.001, significant on paper, marginal in practice.

Step 5: Watch for subgroup spin

When a company reports the “overall” result and then highlights a “pre-specified subgroup” that did better, the subgroup is usually where the drug actually works, and the overall result is soft. Subgroup analyses are underpowered and prone to false positives. The FDA rarely approves drugs based on subgroup analysis of a failed overall trial.

The tell: a press release that says “in the overall population, the trial did not reach statistical significance, but in the pre-specified subgroup of patients with [characteristic], the drug demonstrated [effect].” Read that as: the trial failed, and the company is looking for a salvageable result. For a taxonomy of endpoint-spin patterns with real examples like the Biogen diranersen CELIA trial, see our clinical trial press release reading guide.

Step 6: Read the safety table

Efficacy is half the story. Safety is the other half. Look for three numbers:

Grade 3+ adverse events: life-threatening or severe side effects. Compare the drug arm to the control arm, not to zero.
Discontinuation rate: patients who quit the trial because of side effects. A 15% discontinuation rate in the drug arm versus 3% in control is a real safety signal.
Deaths: more deaths in the drug arm than control is a red flag the FDA will investigate.

A press release that says “safety was consistent with the known profile” without numbers is hiding something. Go to the data table. If the data table is not in the press release, check the company IR page for the conference abstract or NEJM publication.

Step 7: Intention-to-treat vs per-protocol

Intention-to-treat (ITT) analysis includes every patient who was randomized, whether they completed treatment or not. It is the conservative analysis. Per-protocol analysis includes only patients who completed the treatment as specified. It is the optimistic analysis.

If a company leads with per-protocol results and buries ITT, ask why. The FDA prefers ITT because it reflects real-world treatment: patients miss doses, drop out, and switch therapies. Per-protocol can inflate effect sizes by excluding the patients who did poorly.

Step 8: Compare press release to abstract to full publication

The press release is marketing. The conference abstract is a summary. The full publication in NEJM, Lancet, or JAMA is the actual data. What gets buried between the press release and the full paper is where the real risk lives.

Companies time press releases for market hours. Full publications drop at medical conferences like ASCO or ESMO, sometimes weeks later. If a company releases “positive topline Phase 3 data” but has not published or presented the full dataset, the market is trading on a summary. Wait for the full data before you size up.

Real example: Kelun-Biotech and Merck announced the Phase 3 OptiTROP-Lung06 trial of sac-TMT plus Keytruda met its PFS primary endpoint in PD-L1-negative NSCLC. The press release said “statistically significant and clinically meaningful improvement.” Full data is coming at ESMO in October 2026. Our sac-TMT OptiTROP-Lung06 readout coverage notes that the China-only trial design and pending full data are the key caveats. Until ESMO, investors are trading on a summary.

Common mistakes

Trading the headline without reading the primary endpoint. “Positive Phase 3 results” can mean the primary missed and a secondary hit. The stock pops on the headline and fades when the data is digested.
Ignoring the confidence interval. A point estimate that looks great with a CI crossing 1.0 is not statistically significant. This is the most common way investors get fooled by a “positive” readout.
Accepting subgroup analysis as the result. Subgroups are hypothesis-generating, not approval-supporting. If the overall trial failed, the FDA almost certainly will too.
Skipping the safety table. A drug that works but kills 2% of patients gets a CRL, not an approval.
Trading topline before full data. Press-release summaries are incomplete by design. Full data at conferences reveals what the summary omitted.
Forgetting that Phase 3 is the last step before FDA review. Only about 25-30% of drugs that enter Phase 1 make it through Phase 3. A positive readout is necessary but not sufficient for approval. The CRL base rate is 37% of NDAs and BLAs from 2018 to 2022.

Final checklist

Primary endpoint identified and confirmed met or missed
p-value and confidence interval read, not just the point estimate
Effect size compared to clinically meaningful threshold
Subgroup analysis flagged if the overall trial failed
Safety table checked: Grade 3+ AEs, discontinuation rate, deaths
ITT vs per-protocol analysis identified
Full data publication date tracked (ASCO, ESMO, NEJM)
Position sized for binary outcome: 30-50% move in either direction