The New “Peachtree Consensus” for MCED Trials: Cryptic and Vague—or Actually Quite Specific?
The short version
The MCED field is suddenly moving from theory to decisions. Today, September 23, GRAIL’s Galleri test is before an FDA advisory committee, which is considering its PMA for population screening in adults age 50 and older. U.S. Food and Drug Administration
Meanwhile, numerous other blood-based multicancer tests are in development, creating a practical problem: waiting years for cancer-mortality endpoints could make conventional screening trials extraordinarily slow and expensive. The new American Cancer Society-sponsored “Peachtree Consensus” tries to establish rules for using reduction in late-stage cancer incidence as an earlier endpoint. 2026 Cancer Consensus MCED Trials...
See Etzioni et al.
https://pubmed.ncbi.nlm.nih.gov/42775648/
See also: Reboij et al.
https://pubmed.ncbi.nlm.nih.gov/40694037/
I have heard the Etzioni paper characterized as cryptic or vague.
That is only half right. It is actually quite explicit about how an MCED trial should be designed, analyzed, and reported. What it deliberately does not say is the thing regulators, payers, manufacturers, and guideline writers most want to know: How much stage shift is enough?
In other words: the methodology is fairly concrete; the decision rule is not.
What the Consensus is trying to solve
Cancer mortality remains the ultimate goal of screening. The panel says that plainly. But mortality trials take years, whereas an effective screening test should first manifest its effect by preventing cancers from presenting at an advanced stage. The authors therefore conclude that both mechanistic reasoning and prior trial evidence support late-stage incidence as a legitimate primary trial outcome. But they stop short of declaring it a universally validated surrogate for mortality. 2026 Cancer Consensus MCED Tria…
That distinction is central to the whole paper:
Late-stage incidence can be a primary endpoint without being accepted as a proven surrogate for mortality.
That sounds subtle, and perhaps this is where some of the “cryptic” reaction comes from. But conceptually it is not muddled. The panel is saying: we can make an earlier statistically valid measurement; we cannot automatically translate that measurement into lives saved.
Indeed, recent reviews have found a relationship between reductions in late-stage cancers and mortality, but the relationship varies by cancer and is not sufficiently uniform to say that, for example, a 20% reduction in advanced cancers means a 20% reduction in deaths. PubMed
The seven recommendations are surprisingly concrete
The paper ends with a seven-item summary that is much less mysterious than the surrounding prose. 2026 Cancer Consensus MCED Tria… In practical English:
Do not define “late stage” identically for every cancer. Stage III/IV may make sense for ovarian cancer; for colon cancer, stage IV may be the meaningful dividing line; pancreatic cancer might better be divided by resectability. Non-stageable cancers should generally be left out of the primary stage-shift endpoint. 2026 Cancer Consensus MCED Tria… 2026 Cancer Consensus MCED Tria…
Make sure the two trial arms are staged comparably. Otherwise more intensive imaging in one arm could manufacture an apparent difference. The panel recommends a minimum staging protocol, documentation of imaging, biomarkers, biopsies and other work-up, and a target maximum time to final staging. 2026 Cancer Consensus MCED Tria…
Do enough rounds of screening and enough follow-up. Screening initially pulls cancers forward in time and can paradoxically increase observed cancer incidence—including advanced cancers. Too short a trial can therefore make an effective test look ineffective. The authors explicitly recommend erring toward more rather than fewer screening rounds when natural-history information is uncertain. 2026 Cancer Consensus MCED Tria…
Report both the pooled MCED result and the individual cancers. The trial may be powered only for the aggregate endpoint, so individual-cancer results will often be descriptive, but readers need to know which cancers produced the aggregate result. 2026 Cancer Consensus MCED Tria…
Give absolute as well as relative benefit. A 20% relative reduction can mean very different things depending on baseline incidence. The authors specifically point to measures such as the number needed to screen to avert one late-stage diagnosis. 2026 Cancer Consensus MCED Tria…
Translate the stage shift into predicted mortality—but label it as modeling. The predicted mortality calculation should disclose assumptions and inputs and include sensitivity analyses. The panel expressly warns that different models can give different answers and that predicted mortality should contextualize—not be mistaken for proof of—clinical utility. 2026 Cancer Consensus MCED Tria…
Do not wait passively for mortality results. If a trial produces a statistically significant and clinically important late-stage reduction, the panel supports moving into large demonstration, consortium, implementation, and observational studies while the randomized trial continues following mortality. 2026 Cancer Consensus MCED Tria…
None of that is particularly cryptic.
Where the paper really is vague
The missing sentence is something like:
“An MCED test demonstrating at least X% reduction in late-stage cancer, with Y confidence interval and Z safety profile, has shown adequate clinical benefit.”
There is no X, Y, or Z.
The authors acknowledge the issue directly. They say there may be a desire for a single threshold for an adequate late-stage incidence reduction, but that setting one is difficult in a multicancer test. 2026 Cancer Consensus MCED Tria…
That omission is consequential. Earlier MCED work has sometimes used roughly 20% reduction in late-stage incidence as a trial-design benchmark, but the literature itself notes that the magnitude that should count as clinically significant remains unresolved. AACR Journals
Likewise, phrases such as “significant late-stage reduction suggestive of meaningful clinical benefit” sound decisive until one asks who decides what meaningful means. 2026 Cancer Consensus MCED Tria…
So does the conclusion's formulation that complementary studies should begin when the data give “confidence” that the test is likely to produce meaningful clinical utility. 2026 Cancer Consensus MCED Tria…
Those are judgment standards, not quantitative standards.
An additional source of apparent vagueness: “late stage” itself moves
This is perhaps the most intellectually interesting part of Peachtree.
One might have expected a consensus statement to standardize the endpoint: say Stage III + IV cancers. Instead, the panel essentially says that such simplicity would sometimes be scientifically wrong.
A stage III colon cancer is often curable; stage III ovarian cancer behaves very differently. For pancreas, resectability may matter more than the Roman numeral attached to the tumor. The paper therefore substitutes a cancer-specific clinical concept of consequential advanced disease for a mechanically uniform staging rule. 2026 Cancer Consensus MCED Tria…
That makes the guidance less plug-and-play—but arguably more scientifically coherent.
The tradeoff is obvious: the more clinically intelligent the endpoint becomes, the less universal and easily audited the rule becomes.
And it does not say that stage shift is enough for population screening
This is an important guardrail that can disappear in summaries of the paper.
The authors explicitly conclude that late-stage incidence alone is currently insufficient as the basis for a population screening recommendation. A convincing stage shift may justify moving rapidly into implementation-oriented studies while mortality follow-up continues; it does not terminate the evidentiary process. 2026 Cancer Consensus MCED Tria…
That is broadly compatible with the UK National Screening Committee's 2026 position: late-stage incidence is considered the most promising interim endpoint, while mortality, harms, quality of life, resource use, and cancer-specific results remain important to a definitive screening judgment. BMJ
The Galleri problem sitting in the background
The timing makes this much more than an academic exercise.
The NHS-Galleri randomized trial used combined Stage III/IV incidence as its primary endpoint. It did not demonstrate a significant reduction in that aggregate primary endpoint, although GRAIL reported reductions in Stage IV disease, including reductions in later screening rounds. GRAIL The FDA panel considering Galleri today therefore confronts almost exactly the methodological questions Peachtree discusses: What definition of advanced disease matters? How much weight should Stage IV receive? How should effects after repeated screening rounds be interpreted? And what can stage shift legitimately tell us about ultimate mortality benefit?
Peachtree does not provide an escape hatch that converts a missed primary endpoint into a positive trial. But it does argue that Stage III/IV is not necessarily the biologically or clinically optimal endpoint for every cancer, that results should be examined by cancer type and screening round, and that predicted mortality can add context.
That makes the Consensus exceptionally timely.
Bottom line for busy readers
Calling the Peachtree Consensus “cryptic” overstates the problem. Calling it “vague” identifies something real—but only at the most important decision point.
The document is quite specific about trial architecture: cancer-specific definitions of late disease, comparable staging, adequate numbers of screening rounds, aggregate plus cancer-specific reporting, absolute plus relative effects, mortality modeling, and continued post-trial evidence generation.
What it does not provide is a regulatory or guideline threshold for success. It tells researchers how to produce a credible stage-shift result, but not exactly when that result becomes good enough to conclude that an MCED test should be used.
Perhaps the cleanest description is:
Peachtree is a rulebook for measuring MCED benefit, not a rulebook for declaring victory.
And with Galleri literally before FDA today, that unresolved second question is no longer theoretical. U.S. Food and Drug Administration
