What a verified read looks like

Most diligence documents assert. This one shows its work, so the reader can check it rather than take it on faith. Immunome's desmoid programme is a useful case precisely because the headline number is so good: when a result is that clean, the discipline that matters is the one that keeps you from skipping the verification.

The number, and where it comes from

The Phase 3 read is a progression-free survival hazard ratio of 0.16, with a 95% confidence interval of 0.071 to 0.375 -- an 84% reduction in the risk of progression or death.11Immunome, Inc., Form 8-K, 2026-03-16 -- "a statistically significant and clinically meaningful 84% reduction in the risk of disease progression or death." https://www.sec.gov/Archives/edgar/data/1472012/000141588926000193/immunome8k031606.htm (T1 -- regulatory filing) That figure is not quoted from a deck. It is in the company's Form 8-K, which is where a claim like this becomes something a regulator has seen.

The confidence interval is doing quiet work. An upper bound of 0.375 means the pessimistic end of this estimate is still a 62% risk reduction. Point estimates get repeated; intervals are what tell you how much room the estimate has to be wrong in.

The endpoint that is harder to argue with

It is worth being precise about what a hazard ratio is, because it is the number most often quoted and most often misread.

The hazard is the risk of an event -- here, the tumour progressing or the patient dying -- among the people still event-free at a given moment. The ratio compares that instantaneous risk in the treated arm against the control arm, averaged across the trial. So 0.16 says that at any point during follow-up, a patient on varegacestat was running about a sixth of the risk of a patient on placebo. Subtract it from one and you get the 84% risk reduction the filing quotes; they are the same number said twice.

Three things it does not say, all of which get assumed:

• It is not a survival rate. It says nothing about how many patients progressed, only about their relative rate of doing so.
• It is not a duration. It will not tell you how much longer anyone has. Two trials with identical hazard ratios can differ by months in median PFS.
• It assumes the ratio holds steady over time. If a drug's advantage is concentrated early and fades later, a single averaged ratio hides that shape entirely. This is why the curves matter and not just the coefficient.

Which is why the response rate is the useful companion. A hazard ratio compares the shape of two curves; a response rate counts tumours that got smaller. It is a coarser question, and correspondingly harder to dispute.

Figure 1. Confirmed ORR, varegacestat versus placebo, assessed by blinded independent central review (p<0.0001).22Immunome, Inc., Form 8-K, 2026-03-16 -- "an objective response rate of 56% vs. 9% with placebo (p<0.0001), as assessed by blinded independent review." https://www.sec.gov/Archives/edgar/data/1472012/000141588926000193/immunome8k031606.htm (T1 -- regulatory filing) Values are encoded twice -- as bar length and as a printed number -- so no width has to be translated back into a quantity. Drawn by figures/fig-1.py from figures/fig-1.csv.

Blinded independent review is the phrase to look for. It means the assessment was not made by the investigators who knew which arm each patient was on.

An exploratory analysis puts the median best change in tumour volume at -83% for varegacestat against +11% for placebo.33Immunome, Inc., Form 8-K, 2026-03-16 -- "In an exploratory analysis, varegacestat demonstrated a median best change in tumor volume of -83% vs. +11%." https://www.sec.gov/Archives/edgar/data/1472012/000141588926000193/immunome8k031606.htm (T1 -- regulatory filing) Exploratory is the operative word: it was not the endpoint the trial was powered to answer, so it belongs in the picture and not in the argument.

What the record establishes, and what it does not

The safety language in the filing is that varegacestat was "generally well tolerated with a manageable safety profile,"44Immunome, Inc., Form 8-K, 2026-03-16 -- "Varegacestat was generally well tolerated with a manageable safety profile." https://www.sec.gov/Archives/edgar/data/1472012/000141588926000193/immunome8k031606.htm (T1 -- regulatory filing) and that the profile was "consistent with the gamma secretase inhibitor class."55Immunome, Inc., Form 8-K, 2026-03-16 -- "consistent with the gamma secretase inhibitor class." https://www.sec.gov/Archives/edgar/data/1472012/000141588926000193/immunome8k031606.htm (T1 -- regulatory filing) The second clause is the more informative one. A class-consistent profile is a known quantity, which is a different and better thing than an unremarkable one.

What the record does not settle is the subgroup question. The read holds so long as response rates hold across CTNNB1 subgroups; that is the axis on which a clean result of this shape would come apart, and it is the thing to watch rather than a thing to predict.

Why the format is the point

Every number above is one click from the filing it came from, and the chart ships with the script that drew it and the data it drew from. That is the whole claim: not that the analysis is clever, but that it is checkable by someone who was not in the room.

A document that asserts is worth what its author is worth to you. A document that cites is worth what its sources are worth, which is a question you can answer yourself.

Footnotes

1. Immunome, Inc., Form 8-K, 2026-03-16 -- "a statistically significant and clinically meaningful 84% reduction in the risk of disease progression or death." https://www.sec.gov/Archives/edgar/data/1472012/000141588926000193/immunome8k031606.htm (T1 -- regulatory filing) ↩

2. Immunome, Inc., Form 8-K, 2026-03-16 -- "an objective response rate of 56% vs. 9% with placebo (p<0.0001), as assessed by blinded independent review." https://www.sec.gov/Archives/edgar/data/1472012/000141588926000193/immunome8k031606.htm (T1 -- regulatory filing) ↩

3. Immunome, Inc., Form 8-K, 2026-03-16 -- "In an exploratory analysis, varegacestat demonstrated a median best change in tumor volume of -83% vs. +11%." https://www.sec.gov/Archives/edgar/data/1472012/000141588926000193/immunome8k031606.htm (T1 -- regulatory filing) ↩

4. Immunome, Inc., Form 8-K, 2026-03-16 -- "Varegacestat was generally well tolerated with a manageable safety profile." https://www.sec.gov/Archives/edgar/data/1472012/000141588926000193/immunome8k031606.htm (T1 -- regulatory filing) ↩

5. Immunome, Inc., Form 8-K, 2026-03-16 -- "consistent with the gamma secretase inhibitor class." https://www.sec.gov/Archives/edgar/data/1472012/000141588926000193/immunome8k031606.htm (T1 -- regulatory filing) ↩