APBI-IMRT Florence NCT02104895
ForEarly breast cancer post-BCS, pT <25 mm, margins ≥5 mm, age >40
7.7% vs 4.2%
HR 1.57 (95% CI 0.82-3.04), p=0.17
TL;DR15-yr IBTR 7.7% APBI vs 4.2% WBI, HR 1.57 (0.82-3.04) p=0.17; excess driven by new ipsilateral primaries, not true local relapse.
The 5-fraction 30Gy IMRT schedule is what transfers: this is the longest follow-up for that specific PBI regimen, and the 15-yr excess sits in new ipsilateral primaries (5.9% vs 2.7%, p=0.09) rather than local relapse (2.1% vs 1.6%, p=0.75). That distinction is the whole case for keeping PBI in Florence-eligible pts, and it rests on adjudication, not on a powered endpoint.
In a woman over 40 after breast-conserving surgery with a tumour under 25 mm and margins of at least 5 mm, this supports offering 5-fraction PBI as a durable option; it does not extend to node-positive disease, close margins, or younger patients.
The transferable parameter is 30Gy in 5 fractions by IMRT, now with 15-yr follow-up. The excess ipsilateral events are new primaries (5.9% vs 2.7%, p=0.09), not local relapse (2.1% vs 1.6%, p=0.75), which is the distinction that justifies continuing PBI in Florence-eligible pts.
Nothing here moves systemic therapy: distant metastasis 2.7% vs 4.6% and breast-cancer deaths 2.3% vs 3.1% are indistinguishable at 15 years. What matters downstream is the new-primary rate of 5.9% with PBI, which shapes how much ipsilateral surveillance a de-escalated local approach earns.
+2 more figures
10 details
Phase III equivalence trial, 1:1 randomisation, n=520, accrued 2005-2013, median follow-up 15 years. Powered at 80% against a 5-year estimated IBTR of 3% with a 5% equivalence margin. Survival outcomes analysed ITT; toxicity and cosmesis per protocol after 14 withdrawals.
Post-breast-conserving-surgery early breast cancer: pT <25 mm, final surgical margins ≥5 mm, age >40 years. A selected, low-risk population by design.
PBI arm: 30Gy in 5 fractions delivered by IMRT (n=260). WBI arm: 50Gy in 25 fractions plus a 10Gy in 5-fraction tumour-bed boost (n=260).
No endpoint separated the arms at 15 years. The IBTR point estimate favours WBI (HR 1.57, 95% CI 0.82-3.04, p=0.17) but the confidence interval spans unity.
Florence remains the only randomised test of a 5-fraction IMRT PBI schedule with follow-up this long; other external-beam PBI randomisations used different dose and fractionation, so their recurrence rates are not directly interchangeable with this one. The direction here, a numerically higher IBTR concentrated in new primaries, is the pattern PBI trials have consistently reported when they separate the two.
The trial was sized for 5-year equivalence, so the 15-year IBTR comparison is a long-term description rather than a powered test, and the upper CI bound of 3.04 leaves room for a real excess. The relapse-versus-new-primary split is an adjudicated distinction, not a molecularly confirmed one, and it carries the entire reassurance.
The result supports continuing PBI in Florence-eligible pts rather than expanding the indication. It does not settle whether the new-primary excess is a genuine consequence of leaving untreated breast tissue unirradiated, which is biologically the expected cost of the approach and would not be captured by any local-control endpoint.
CONSORT flow
Randomised phase III with mature 15-yr follow-up; no significant difference on any oncological endpoint. Supports an already guideline-listed de-escalation rather than establishing a new one.
- Whether the new-primary excess reflects untreated ipsilateral breast tissue
- Ipsilateral surveillance intensity after partial-breast irradiation
- Applicability of 5-fraction PBI below age 40
📚 Sources · 🐦 1 tweet
📌 Fifteen-year outcomes of the randomised APBI-IMRT Florence phase Ill trial of partial versus whole-breast irradiation in early breast cancer ✨
— Elisabetta Bonzano MD, PhD (@to_be_elizabeth) May 17, 2026
Excellent presentation led by @CarlottaB 👏🏻#ESTRO26 @Icro_Meattini @ESTRO_RT @OncoAlert #OncoAlertAF pic.twitter.com/1j4bIA2nyC
The longer read
Fifteen years is long enough to answer the question partial-breast irradiation has always faced: does sparing the rest of the breast eventually cost you, and if so, in what currency. Florence's answer is that the cost, if it exists, is not local relapse. True local recurrences were 5 (2.1%) with PBI and 4 (1.6%) with whole-breast, which is as close to identical as this trial size can resolve. The entire numerical gap in ipsilateral events sits in new primaries, 15 (5.9%) versus 7 (2.7%), p=0.09. That is exactly the failure pattern the mechanism predicts. Whole-breast irradiation sterilises occult disease throughout the breast, including clones that would surface a decade later as apparently new cancers; partial-breast irradiation does not, and by design was never meant to. Reading the composite IBTR as though it were a local-control failure conflates two different events with two different clinical consequences.
That framing has a limit worth stating plainly. The trial was powered on a 5-year IBTR estimate of 3% with a 5% equivalence margin, so the 15-year comparison is descriptive follow-up, not a test the trial was built to pass. The HR of 1.57 with an interval reaching 3.04 is compatible with equivalence and also compatible with a doubling of ipsilateral events. A reader who wants Florence to prove no difference is asking more of it than it can give; what it does give is fifteen years without a signal on any endpoint that carries survival implications, with locoregional recurrence at 7.2% versus 5.0% (p=0.28), distant metastasis 2.7% versus 4.6%, breast-cancer deaths 6 versus 8. Nothing accumulated in the direction that would make an oncologist reconsider.
The relapse-versus-new-primary adjudication deserves scepticism because it is load-bearing. This is a clinicopathologic judgment, not a molecular one, and the classification of an ipsilateral event as new rather than recurrent shifts it from the column that indicts the technique to the column that exonerates it. The distinction is standard and defensible, but an analysis whose reassuring conclusion depends on one adjudicated variable should be read with that dependency in view. A reader who declined the distinction entirely would be left with a non-significant numerical excess in a trial not powered to detect it, which is a weaker claim but not a contradictory one.
What actually transfers to practice is narrower than the headline. The tested regimen is 30Gy in 5 fractions by IMRT against 50Gy in 25 plus a 10Gy boost, in women over 40, after breast-conserving surgery, with tumours under 25 mm and margins of at least 5 mm. Those eligibility criteria are not incidental; they select a population whose absolute recurrence risk is low enough that a modest relative difference stays clinically small. Extending the schedule to larger tumours, closer margins, node-positive disease or younger women asks it to perform where the absolute risk is higher and the arithmetic of a relative excess is less forgiving. Florence is best read as durable support for keeping a de-escalation option already in use, in the patients it was tested in, rather than as licence to widen it.