EORTC IM-MS (22922/10925)
ForStage I-III breast cancer considering internal mammary / medial supraclavicular nodal RT
61.0% vs 61.8%
HR=1.00, 95% CI 0.90-1.10, P=0.967 (1° EP not met)
TL;DR20yr OS 61.0% vs 61.8%, HR 1.00 (0.90-1.10), P=0.967: the 15yr breast-cancer mortality benefit is erased by non-BCM.
The 20yr null OS is a competing-risk cancellation, not absent efficacy: BCM 18.6% vs 22.4% (HR 0.82) offset by non-BCM 20.4% vs 15.8% (HR 1.26), with cardiac disease 15.2% vs 11.7% and lung fibrosis 6.3% vs 3.2%. Whether IM coverage survives depends entirely on your heart dose, and DBCG IMN2 ran 4-9x lower.
In stage I-III breast cancer where IM-MS coverage is on the table, this supports the anti-cancer effect of IM irradiation while showing the 20yr survival gain is forfeited at 1990s-era heart doses; it does not describe outcomes at modern DIBH/IMRT cardiac exposures.
This is a planning-constraint result, not a target-volume result. IM coverage delivered BCM 18.6% vs 22.4% (HR 0.82) and gave it all back as non-BCM (HR 1.26), with cardiac disease 15.2% vs 11.7%. The benefit survives only if your mean heart dose looks like DBCG IMN2's 1.2-2.3 Gy, not this trial's.
The 20yr OS null (61.0% vs 61.8%) should not be read as breast-cancer control being unimproved: BCM fell to 18.6% from 22.4%. The excess deaths are cardiopulmonary, which is a comorbidity and cardiac-surveillance consideration in long survivors treated in older RT eras rather than a systemic-therapy signal.
| Endpoint | IM-MS RT | No IM-MS RT | HR | P |
|---|---|---|---|---|
| BCM rate | 18.6% | 22.4% | 0.82 (0.72-0.95) | 0.006 |
| non-BCM rate | 20.4% | 15.8% | 1.26 (1.09-1.46) | 0.002 |
+3 more figures
| Endpoint | IM-MS RT | No IM-MS RT | HR (95% CI) | P |
|---|---|---|---|---|
| Overall survival (ITT), 20yr | 61.0% | 61.8% | 1.00 (0.90-1.10) | 0.967 |
| RT-related side effect | IM-MS RT | No IM-MS RT |
|---|---|---|
| Lung fibrosis | 6.3% | 3.2% |
| Cardiac fibrosis | 2.7% | 1.7% |
| Cardiac diseases | 15.2% | 11.7% |
| Endpoint (pN0) | IM-MS RT | No IM-MS RT | HR | P |
|---|---|---|---|---|
| DFS rate | 53.9% | 53.6% | 0.93 (0.81-1.07) | 0.318 |
| DMFS rate | 67.2% | 67.4% | 0.93 (0.78-1.10) | 0.397 |
9 details
Randomised EORTC trial 22922/10925, stage I-III breast cancer, internal mammary + medial supraclavicular irradiation versus no IM-MS irradiation. This is the 20-year readout, presented as a plenary at ESTRO 2026, including a dedicated pN0 analysis.
The intervention is the IM-MS target volume itself, added to otherwise standard locoregional treatment. Per-arm dose and fractionation are not reported in source; what the presenters did quantify is the dosimetric era gap, with DBCG IMN2 mean heart doses 4-9 times lower (MHD 1.2 Gy right-sided, 2.3 Gy left-sided, treated 2007-2014).
Primary: overall survival (ITT). Secondary readouts presented at 20 years include DFS and DMFS under DATECAN definitions (DFS counts all deaths and all breast events including DCIS and contralateral; DMFS counts all deaths and distant metastases), breast-cancer mortality, non-breast-cancer mortality, second cancers, and RT-related late effects.
OS 61.0% vs 61.8%, HR=1.00 (0.90-1.10), P=0.967. DFS and DMFS are likewise flat (HR 0.97 each), and the pN0 subgroup shows no separation. The signal lives entirely in the cause-specific split.
No statistical difference in secondary cancers or second breast cancers between arms. Absolute RT-related late effects favour the control arm: cardiac disease 15.2% vs 11.7%, lung fibrosis 6.3% vs 3.2%, cardiac fibrosis 2.7% vs 1.7%.
The Danish DBCG IMN2 cohort (Nielsen, Lancet Reg Health Eur 2024) reported that IM-MS irradiation reduced distant metastasis and BCM and improved OS in node-positive pts at 15 years. The presenters attribute the divergence to technique, since IMN2 heart doses were 4-9x lower.
The pN0 result is a subgroup read on endpoints that were flat overall, so it cannot exclude a small benefit in that group. The competing-risk interpretation also rests on comparing this trial's toxicity against a non-randomised cohort treated a decade later, which is a dosimetric argument, not a trial result.
This is the cleanest available demonstration that an oncologically real nodal-RT benefit can be spent entirely on late cardiopulmonary mortality. It settles that IM-MS irradiation reduces breast cancer death; it does not settle whether the survival benefit is recoverable, which is now a planning question rather than a target-volume question.
Randomised, prespecified 1° OS, 20yr follow-up, null. Divergence from DBCG IMN2 is confounded by an old-technique heart dose (4-9x higher), not by design flaw.
- Does IM-MS survival benefit re-emerge at modern cardiac-sparing doses?
- Which nodal subgroups justify IM coverage given competing cardiac mortality?
📚 Sources · 🐦 2 tweets
📌 Internal Mammary and Medial Supraclavicular irradiation in stage I-III breast cancer: 20 years results of the randomised EORTC trial 22922/10925, including in pNo patients
— Elisabetta Bonzano MD, PhD (@to_be_elizabeth) May 17, 2026
Special Joint Presentation Led by Prof. Philip Poortmans and Orit Kaidar-Person ✨ at #ESTRO26 @ESTRO_RT… pic.twitter.com/KIoJtdhEzp
20-year outcomes of @EORTC internal mammary #radiotherapy trial.
— Shankar Siva (@_ShankarSiva) May 17, 2026
➡️internal mammary improved control
➡️ survival counterbalanced by late adverse events #radiotherapy #bcsm
Great to see the long term data at #ESTRO26, and discussing Charlotte Cole suggests with modern RT, long… pic.twitter.com/yPtlfrLcri
The longer read
The most useful thing about this readout is that it does not show IM-MS irradiation failing to work. It shows it working, and then shows the winnings handed back. Breast cancer mortality at 20 years is 18.6% versus 22.4%, HR 0.82 (0.72-0.95), P=0.006, which is a real and statistically clean anti-cancer effect on the endpoint the target volume was drawn to influence. Non-breast-cancer mortality moves the other way, 20.4% versus 15.8%, HR 1.26 (1.09-1.46), P=0.002, and the two forces meet almost exactly at the primary endpoint: OS 61.0% versus 61.8%, HR 1.00, P=0.967. A hazard ratio that lands on unity with a confidence interval of 0.90-1.10 after 20 years and roughly four thousand patients is not an underpowered null. It is a genuine cancellation.
That framing matters because the two hazard ratios sit on different footings for a reader deciding what to do tomorrow. The BCM effect is a property of covering the internal mammary chain, and there is no obvious reason it should decay with better planning. The non-BCM excess is a property of what the beam did to the heart and lung on its way there, and that is exactly the part of the trial that has been engineered against for two decades. The late-effect rates support the mechanism rather than leaving it as speculation: cardiac disease 15.2% versus 11.7%, cardiac fibrosis 2.7% versus 1.7%, lung fibrosis 6.3% versus 3.2%. Second cancers, notably, did not differ, so the excess mortality is not being carried by radiation-induced malignancy.
The presenters' own comparator does the rest of the work. DBCG IMN2 (Nielsen, Lancet Reg Health Eur 2024) reported reduced distant metastasis, reduced BCM and improved OS in node-positive patients at 15 years, and its mean heart doses were stated as 4 to 9 times lower than in EORTC IM-MS, with MHD 1.2 Gy for right-sided and 2.3 Gy for left-sided treatment in a 2007-2014 cohort. Two trials of the same target volume reaching opposite survival conclusions, with an order-of-magnitude difference in the organ-at-risk dose that plausibly explains the difference, is about as close to a mechanistic reading as observational comparison allows. It is still observational. IMN2 was not randomised against EORTC, the eras differ in systemic therapy as well as in planning, and node-positive selection is not the same population as this trial's ITT set.
The pN0 analysis deserves less weight than its billing. DFS 53.9% versus 53.6% (HR 0.93, 0.81-1.07) and DMFS 67.2% versus 67.4% (HR 0.93, 0.78-1.10) are flat on endpoints that were flat in the whole trial, and the confidence intervals are wide enough to be compatible with a modest benefit or a modest harm. What it does establish is that nothing in the pN0 data argues for extending IM-MS coverage into a group where the competing-mortality tax would apply in full and the absolute recurrence risk is lowest. That is the group where the technique argument helps least, because a smaller anti-cancer effect is being purchased with the same cardiac exposure.
What would have to be true for the optimistic reading to be wrong: that some part of the non-BCM excess is not dose-driven and therefore not recoverable by planning, or that the BCM benefit itself was inflated by the systemic therapy of the enrolment era and shrinks against contemporary backbones. Neither can be excluded from these data. The honest conclusion is narrow. IM-MS irradiation reduces death from breast cancer, at 1990s-era cardiac dose it does not extend life, and whether it does so now is a dosimetry question that this trial cannot answer.