How to read clinical trial results

A Phase 3 readout is the single most consequential event in most biotech investment theses. It is also where the gap between what a press release says and what the data shows is widest. "Met its primary endpoint" can describe a practice-changing result or a statistically significant sliver that no payer will reimburse. Learning to tell the difference is the whole game.

This is a framework for reading a trial result the way a diligence analyst does. It is written around a Phase 3 topline release, for the reason above, but the same checks apply to an earlier-stage readout. It is not a recommendation about any specific company. Nothing here is investment advice.

Start with the question the trial was designed to answer

Before you look at a single number, reread the trial design. A readout only means something relative to what was promised months or years earlier on ClinicalTrials.gov, where the primary outcome measure is a required registration element.

• What is the primary endpoint, exactly? Overall survival is a higher bar than progression-free survival, which is a higher bar than a response rate. The latter two are surrogates, and under FDA's accelerated pathway a surrogate need only be "reasonably likely" to predict clinical benefit. A change in the primary endpoint late in a trial's life is a yellow flag: ICH E9 calls redefinition after unblinding "almost always ... unacceptable".
• What was the statistical plan? The pre-specified alpha, the powering assumptions, and any interim analyses, which E9 requires be planned in advance, tell you what the company expected the effect size to be. A trial powered for a large effect that squeaks by on a small one is a different outcome than the design anticipated.
• What is the comparator? Superiority over placebo is not the same as superiority over the current standard of care, or non-inferiority to it. The commercial question is almost always "versus what patients actually get today."

Separate statistical significance from clinical meaning

A p-value tells you whether an effect is likely real. It tells you nothing about whether the effect is large enough to matter. The American Statistical Association says this outright: statistical significance "does not measure the size of an effect or the importance of a result." These are two independent questions.

Look for the effect size and its confidence interval, not just the p-value:

• A hazard ratio of 0.65 with a tight confidence interval is a strong, precise signal.
• A hazard ratio of 0.85 with an interval that nearly touches 1.0 can be "statistically significant" and still commercially underwhelming.
• Ask whether the magnitude clears the bar that regulators, guideline committees and payers care about. That bar sits above p < 0.05, and ESMO grades the magnitude explicitly with its Magnitude of Clinical Benefit Scale.

Read the secondary endpoints and subgroups carefully

Secondary endpoints that move in the same direction as the primary reinforce a result. Secondaries that miss complicate the story and often the label. Watch in particular for a missed overall-survival secondary sitting under a hit progression-free-survival primary.

Subgroup analyses deserve particular skepticism. A prespecified subgroup with a biological rationale is evidence. A subgroup discovered after the fact to rescue a failed primary is a hypothesis, not a result. ICH E9 holds that conclusions based solely on exploratory subgroup analyses are "unlikely to be accepted," and the EMA has a guideline on the distinction. The phrase "benefit was driven by" should always prompt the question: was that subgroup defined before or after the data unblinded?

Never skip the safety table

Efficacy gets the headline; safety writes the label. Scan for:

• Discontinuation rate due to adverse events. The cleanest proxy for real-world tolerability. ICH E3 requires reports to break discontinuations out by reason.
• Serious adverse events and deaths, and how they balance against the comparator arm.
• Any signal that could trigger a boxed warning, a REMS program, or a narrower indication than the trial population.

A drug that works but is hard to tolerate competes very differently than the efficacy number alone suggests.

Put the readout in its commercial context

Once you understand the data, the investment-relevant questions are about what the data enables:

• Does this result support the label the thesis assumed, or a narrower one?
• How does the effect size compare to already-approved competitors and to anything else in late-stage development for the same indication?
• Does it change the probability of approval, the addressable population, or the pricing power?

The topline is the beginning, not the end

A topline press release is a curated summary written by the sponsor. The fuller picture arrives later: at a medical conference, in a peer-reviewed publication, and in the FDA's own review documents. Hold your conclusions loosely until then. The topline gave you the sponsor's summary of the numbers; the conference and the review documents give you the numbers. For what a topline release does and does not contain, and roughly how long the rest takes to arrive, see what does topline data mean.

That is exactly the work FuzeBio is built to accelerate: pulling the trial design, the endpoints, the competitive set, and the regulatory history into one place so you can judge a readout on its merits instead of its headline.