KEYNOTE-942: what the original melanoma vaccine trial found

The original KEYNOTE-942 paper compared an individualized cancer vaccine plus pembrolizumab with pembrolizumab alone after melanoma surgery. It was a randomized Phase 2b trial, not the later INTerpath-001 Phase 3 trial. This page explains the historical analysis published in The Lancet in 2024. 1

For the current program milestone, go to the 2026 intismeran report. Keeping this page historical allows the numbers to remain tied to their actual follow-up period.

The original comparison in one place

The study randomized 157 people with completely resected stage IIIB–IV cutaneous melanoma: 107 to combination treatment and 50 to pembrolizumab alone. It was open-label. Median follow-up was approximately 23 and 24 months in the two groups. 1

The abstract reported:

These figures are from the accessible trial abstract, not a new calculation or the 2026 Phase 3 results. The investigators concluded that the results supported potential benefit from individualized neoantigen therapy in this setting. 1

Why 79% versus 62% is not a 79% cure rate

The time point and endpoint are part of the result. An estimate of remaining recurrence-free at 18 months does not show permanent freedom from disease. Nor does it describe everyone with melanoma, irrespective of stage, previous treatment or whether the cancer has been surgically removed.

The arithmetic difference between those rounded point estimates is 17 percentage points. That subtraction is not the hazard ratio. It also does not reproduce a confidence interval for the difference or establish a universally applicable number needed to treat.

A hazard ratio summarizes a relative rate over follow-up under a statistical model. A time-specific survival estimate answers a different question. Using both can be informative; silently substituting one for the other makes an apparently simple headline misleading.

What the confidence interval tells the reader

The original reported interval extended slightly above 1, and its two-sided p-value was slightly above 0.05. Preserve those facts rather than turning the point estimate into certainty. 1

It would also be an overcorrection to say the study therefore demonstrated “no effect.” The interval expresses uncertainty around the estimate. Interpreting a trial requires its design, prespecified analysis and accumulated evidence, not only whether one displayed p-value falls on one side of a familiar cutoff.

This report does not recreate the study’s full statistical analysis plan. The accessible primary abstract is sufficient to show why “proved a precise permanent risk reduction” is too strong a summary of this original analysis.

Combination evidence is not vaccine-alone evidence

Both groups received pembrolizumab. The additional experimental component was the individualized vaccine. The comparison therefore helps evaluate the added regimen; it does not tell us what would happen if the vaccine replaced every other component of care. 1

That distinction is especially important when a reader searches for an alternative to an existing cancer treatment. The studied question may have been addition, not substitution. The immune-therapy comparison explains the different roles without ranking unlike treatments.

Later results do not rewrite the date on this analysis

A later follow-up can contain more events and a revised effect estimate. A separate Phase 3 trial can test a related hypothesis in a larger or different population. Neither means that this historical page should silently replace its original numbers with the latest percentage found in a press release.

The source trail should remain: original trial, subsequent analysis of that trial, then independently identified later study. That makes apparent changes interpretable rather than confusing.

Is KEYNOTE-942 accepting new participants?

The NCI record for NCT03897881 was marked closed to accrual when checked on September 17, 2026. The record is still useful for identifying the study even when it is not enrolling new participants. 2

A trial having a new paper does not imply that enrollment reopened. For a different study, use its own identifier and site contact; do not carry the status from one protocol to another. Our cancer-vaccine trial guide explains that search process and its limits.