EORTC 22922/10925
ForStage I-III breast, central/medial tumor or involved axilla, post-ALND
61.0% vs 61.8% at 20yr
HR 1.00, p=.967; primary endpoint not met
TL;DR20yr OS 61.0% vs 61.8% (HR 1.00, p=.967): IM-MS nodal RT cut breast cancer mortality but added non-cancer deaths.
The breast cancer mortality gain (22.4% vs 18.6%, HR 0.82) is real and was fully offset by non-breast-cancer deaths (15.8% vs 20.4%, HR 1.26) that only emerged after 15 years. Cardiac disease ran 15.2% vs 11.7%. In a 1996-2004 planning era, that trade gates IM-MS coverage on achievable heart dose, not on nodal risk alone.
In a woman with a medial or central stage I-III tumor being considered for IM chain coverage, this supports treating the cardiac dose constraint as co-equal with nodal risk; it does not speak to modern DIBH or proton delivery, where the competing-mortality arm may not hold.
Breast cancer mortality fell (HR 0.82) and non-cancer death rose (HR 1.26), netting OS HR 1.00, with cardiac disease 15.2% vs 11.7% and lung fibrosis 6.3% vs 3.2%. Planning ran in the 2D era, so the decision this moves is achievable heart dose, not whether to cover the IM chain.
Systemic therapy was per institutional preference across 1996-2004 accrual, so the disease-specific gain (HR 0.82) sits on a backbone that predates current regimens. For a med onc the read is competing mortality: the excess non-cancer deaths after 15 years reframes how long-term cardiac surveillance should run in irradiated survivors.
9 details
Prospective multicenter randomized phase 3 trial, accrual 1996-2004, 4004 pts, with an RT quality-assurance program built in. The last analysis was planned at 20 years on the assumption that any survival effect of IM-MS-RT would be delayed. Median follow-up 22.2 years.
Women ≤75 yrs, unilateral histologically confirmed breast adenocarcinoma, stage I-III. Gate was tumor location or nodal status: centrally or medially located primary irrespective of axillary involvement, or any quadrant with axillary involvement. Median age 54.
Randomization was to internal mammary and medial supraclavicular (levels 3-4) nodal irradiation or not, layered on standard breast or chest wall treatment. Dose and fractionation are not given in the source excerpt. Planning ran in the two-dimensional and early-conformal era, when IM coverage at least doubled heart dose.
Primary: overall survival. Secondary: disease-free survival, distant metastases-free survival, breast cancer mortality, any breast recurrence.
Lung fibrosis 6.3% vs 3.2%, cardiac fibrosis 2.7% vs 1.7%, cardiac disease 15.2% vs 11.7% with IM-MS-RT. Severe (grade 3-4) events were uncommon and near-equal: cardiac 1.9% vs 1.7%, lung 0.3% vs 0.0%, so the excess sits in lower-grade, chronic morbidity rather than catastrophic events.
MA.20 and DBCG-IMN both read regional nodal RT positively at roughly ten-year horizons, and DBCG-IMN reported an OS gain. This trial covers the same anatomic question at twice that follow-up and shows the disease-specific gain surviving while the survival gain does not, which is the read those trials were too short to produce.
Systemic therapy was left to physician and institutional preference across an eight-year accrual, so the systemic backbone is heterogeneous and predates current regimens. The competing-mortality signal is a cause-of-death attribution over two decades in a population whose baseline cardiovascular risk rises independently, and the source does not report a cardiac-specific mortality breakdown separating RT-attributable from age-attributable death.
The two effects are both real and point opposite ways: HR 0.82 on breast cancer mortality, HR 1.26 on other deaths, netting HR 1.00 on OS. That arithmetic is the finding, and it argues the relevant question is not whether IM-MS coverage works but whether its cost can be engineered down.
| Endpoint | Control | IM-MS-RT | Effect size |
|---|---|---|---|
| Overall survival | 61.8% | 61.0% | HR 1.00, p=.967 |
| Disease-free survival | 49.0% | 48.2% | HR 0.97 (0.89-1.06), p=.5148 |
| Distant metastasis-free survival | 59.8% | 58.9% | HR 0.97 (0.88-1.08), p=.578 |
| Breast cancer mortality | 22.4% | 18.6% | HR 0.82 (0.72-0.95), p=.006 |
| Death not from breast cancer/unknown | 15.8% | 20.4% | HR 1.26, p=.002 |
Randomised, prespecified 20yr primary analysis, adequate power, endpoint reported honestly. Divergence from the 10yr-era read of nodal RT is internally valid, not a design artifact.
- Does modern heart-sparing delivery erase the excess non-cancer mortality?
- Which subgroups have enough breast cancer risk to justify the trade?
- Cardiac surveillance interval for irradiated long-term survivors
📚 Sources · 📄 1 paper
Abstract
The longer read
The result that matters here is not the null overall survival curve but the two opposing hazards underneath it. Irradiating the internal mammary and medial supraclavicular chains reduced breast cancer mortality (HR 0.82, p=.006) and reduced any breast recurrence (HR 0.88, p=.0369). It also raised death from other or unknown causes (HR 1.26, p=.002). Those two are close enough in size that overall survival lands at HR 1.00. A trial that reported only the primary endpoint would look like a clean negative; a trial that reported only the disease-specific endpoint would look like a win. Both readings are wrong, and the reason this paper is worth a radiation oncologist's time is that it is one of very few datasets long enough to show the crossing.
Timing is the crux. The authors state the excess in non-breast-cancer deaths appeared after 15 years. Regional nodal irradiation trials that read out at roughly a decade, MA.20 and DBCG-IMN among them, were structurally incapable of seeing this. Their positive readings were not wrong for their follow-up; they were incomplete. That is an argument about the evidence base for nodal RT generally, not about this trial's arm assignment, and it is why the authors close by calling for very long follow-up in a disease with this prognosis. It also cuts the other way: a benefit that requires twenty years to be offset is a benefit for a patient whose competing risks are low.
How much confidence the toxicity arm should carry is the genuinely contestable part. Accrual ran 1996 to 2004. The paper's own introduction notes that including the IM nodes at least doubled heart dose in the older literature, and that match-line fibrosis at the junction between IM and breast fields was a planning-era problem. Deep inspiration breath hold, IMRT, and routine cardiac dose constraints were not standard practice for most of this cohort. A reader who plans IM coverage today with a documented mean heart dose is not delivering the intervention this trial randomized, and the 1.26 hazard on non-cancer death is the number most likely to be smaller in their hands. Whether it goes to unity is unknown and this trial cannot answer it.
Against that, two cautions. The excess morbidity here is chronic and low-grade rather than catastrophic: severe cardiac events were 1.9% versus 1.7% and severe lung events 0.3% versus 0.0%, essentially equal, while cardiac disease overall was 15.2% versus 11.7%. A mechanism that operates through modest chronic injury over two decades is harder to engineer away than one driven by rare high-dose events, and dose-sparing may compress it rather than remove it. And cause-of-death attribution across twenty years in a cohort aging from a median 54 is a noisy instrument, with the non-breast-cancer category explicitly pooling unknown causes.
The practical read is that IM-MS coverage buys disease control at a cardiopulmonary price, and the decision belongs at the planning stage rather than the indication stage. In a patient whose anatomy permits low heart dose, the mortality arithmetic favors coverage; where it does not, this trial says the trade may not be worth taking.