What we’ll likely see in Merck and Moderna’s upcoming cancer vaccine data
In August, journalists and investors went gaga over Merck and Moderna’s claim of “positive topline results” for their Phase 3 trial combining Keytruda with a patient-specific cancer vaccine. Despite the lack of disclosed details, STAT’s Adam Feuerstein called the outcome of INTerpath-001 “stunning” and cited it as a reason why “2026 will go down as one of the most consequential years in cancer treatment, ever.” And both companies’ stock prices soared on the news, presumably because the press release removed the overhanging risk of an outright failure.
But in the wake of the skimpy press release, now the question shifts to: How positive? Specifically, how will the combo’s Phase 3 efficacy compare to what was seen in Phase 2?
Going into the companies’ presentation of more details at the ESMO conference next week, I think there’s a decent chance the combo will confirm the Phase 2 results, but there are reasonable odds the efficacy will be less impressive, and I’ll be very surprised if the results outperform the prior trial. In the rest of this post, I’ll explain my reasoning.
First, some quick background and context:
For the Phase 3 study, INTerpath-001, the main sources of info are its clinicaltrials.gov record (NCT05933577) and the sponsors’ 8/19/26 press release.
We also have data from a completed Phase 2b study, KEYNOTE-942: initial results (including a detailed supplement) were published in Lancet in 2024, and updated results with 2 more years of follow-up (5 total) were presented at the ASCO conference in 2026.
The two trials share some similarities: Both compared a combination of the vaccine and Keytruda to Keytruda alone and randomized patients 2:1 to the combo vs. monotherapy. And both had the same primary endpoint, recurrence-free survival (RFS), which is the length of time until the cancer recurs/spreads or the patient dies.
There are two notable design differences between the studies based on what’s been disclosed, however: the study size and the stage of the eligible patients’ cancers. KEYNOTE-942 enrolled 157 patients (104 drug-treated) with resectable stage IIIB-IV melanoma; that means patients had lymph node involvement and/or metastatic disease but all identifiable tumor deposits were surgically removed. INTerpath-001 was larger (1,137 patients; est. 758 drug-treated) and expanded eligibility to also allow patients with resectable stage IIB/C disease (deeper tumors with or without ulceration, but no lymph node involvement or metastatic disease) to enroll, in addition to those with resectable stage IIIB-IV disease.
When Merck and Moderna said INTerpath-001 was “positive” in their press release, they didn’t provide any additional information. Based on what’s been disclosed and what’s known overall about drug development, I think there are good reasons to expect equivalent or slightly worse results than KEYNOTE-942.
As context, in the first 18-month analysis of KEYNOTE-942 – analogous to what will be disclosed next week for INTerpath-001 – the combo had an HR for RFS of 0.561 with a 2-sided P value that just missed statistical significance (0.053). So, if INTerpath-001 were basically identical to KEYNOTE-942 except for its larger size, we might expect basically “par” results: an effect size of roughly 0.561 but a tighter 95% CI (smaller P-value) at the early time point due to the higher statistical power.[1]
But there are two reasons why we might expect to see less impressive efficacy in INTerpath-001. A general one is that efficacy decreases are common in Phase 3 studies after positive Phase 2 trials. That’s been seen overall, in solid tumors, and in cancer immunotherapy specifically. Some of that is due to reversion to the mean, but operational factors (due to expanding to many more sites/countries) and broader inclusion criteria may also play a role.
The second reason we might expect to see lower efficacy in INTerpath-001 than KEYNOTE-942 is directly related to the boost in sample size compared to the prior study. All else being equal, a larger study can detect a smaller effect than a smaller one. It’s hard to reconstruct the power calculation for INTerpath-001 from the scant public information, but it’s likely that the threshold for “positive” is somewhat lower given that there was a 7-fold upsize from Phase 2 to Phase 3.
Now, let’s examine the other side: Are there any plausible reasons to expect a “bull case” where INTerpath-001 might overdeliver versus KEYNOTE-942?
The main one I can think of is related to the inclusion criteria – but this has some caveats.
It’s not crazy to think the shift in eligibility from Phase 2 to Phase 3 is a net positive for the combo. As noted above, KEYNOTE-942 required patients to be in stage IIIB or higher, whereas INTerpath-001 allowed earlier-stage patients to join the study. If cancer vaccines work by activating the patient’s immune system to destroy occult tumor cells, then patients with less tumor burden might be more likely to have a significant response.
But the shift toward earlier stages could also cut the other way, because those patients will respond better to Keytruda alone, which raises the bar for the combo to show a statistically signficant improvement. We already saw in a prior study, KEYNOTE-716, that patients with resected stage IIB/IIC melanoma had a median RFS at five years of 67.1% on Keytruda alone, far higher than the 49.1% 5-year mRFS in the Keytruda-only control arm of KEYNOTE-942.[2]
On balance, if INTerpath-001’s patients are mostly later-stage, then both effects are probably small and a wash: the mechanism argument won’t contribute much on the positive side, but the control arm won’t significantly over-perform compared to the prior study. But if there’s a sizable fraction of earlier-stage patients in the Phase 3 trial, both opposing effects could be magnified, and it’s impossible to predict how they will net out.
Finally, I’ve seen one other “bull case” argument on X, but I don’t think it’s valid. Although Merck and Moderna said the “positive” results were observed at an interim analysis, it’s false to conclude this is evidence that INTerpath-001 will show higher efficacy than KEYNOTE-942. All we learned from the successful interim analysis was that the effect is greater than the minimum difference the full trial was powered to detect. But as discussed above, we don’t know what that threshold is. If INTerpath-001 was powered to yield a positive outcome even with an RFS HR of, say, 0.4, then the positive interim result could be better than that but still worse than the 0.561 result seen in KEYNOTE-942. The positive interim result is uninformative without knowing how the trial was powered, and the press release does not provide any objective evidence that the Phase 3 study will outperform what was seen in Phase 2.
My point here isn’t to claim the Merck/Moderna combo isn’t a big deal — it is! Until now, the cancer immunotherapy field has struggled to deliver improved results through combinations with anti-PD-1 or anti-PD-L1 drugs. Regardless of whether the combo replicates the Phase 2 results or falls slightly short, it will further improve the lives of some melanoma patients, which is great news. Merck and Moderna deserve our thanks for taking on the risk and expense to move this work forward.
But a key aspect of the biotech business involves making smart, data-driven predictions about clinical trial outcomes, and this is a useful teaching case in that regard. Yes, Merck and Moderna’s announcement that INTerpath-001 was positive took some downside off the table, but it didn’t tell us anything about the magnitude of the combo’s efficacy. As outlined above, I think the most likely outcome here is par to slightly below-par performance relative to KEYNOTE-942. We’ll see in a few days.
[1] For these sorts of "event studies", the power is driven by the total number of events (here, recurrences or deaths) seen in a given time. That's why the stats can tighten up with either the passage of time (as more events accrue) or a larger sample size (which can lead to more events in the same time).
[2] Notably, although KEYNOTE-942 permitted stage IIIB patients to enter, the published paper shows that all the patients who actually enrolled were later-stage (IIIC or higher). This further reduced the monotherapy (control arm) response in KEYNOTE-942 versus what might have been expected.