onc brain

The longer read

The headline association, that locoregional failure after breast-conserving therapy marks a worse outcome, is not new; the authors place it alongside the NSABP B-06 recurrence analyses, and the EBCTCG overview cited here argued that local control and breast cancer mortality are linked. What this pooled cohort adds is the node-positive, chemotherapy-treated setting and a clean separation of in-breast failure from regional and other locoregional failure. Those two behave as different diseases. An IBTR roughly doubles the mortality hazard (HR 2.58), while an oLRR carries an HR of 5.85, and the gap in 5-yr OS after the event (59.9% vs 24.1%) is too wide to read as one phenomenon.

Methodologically, modeling recurrence as a time-varying covariate is the right choice. It avoids crediting the recurrence-free group with the survival time pts needed simply to be at risk of recurring, a bias that inflated older landmark-free comparisons. Cumulative incidence with competing risks is also appropriate. But the design answers a prognostic question, not a causal one. oLRR clusters with nodal burden and ER and PgR negativity, the same features that drive distant failure, so the recurrence may be the first visible sign of disseminated disease rather than a source of it. Nothing here apportions how much of the excess mortality preventing the recurrence would avert, and a reader should not take the HRs as the survival gain available from better local therapy. The timing data support the same caution from the other direction: later IBTRs carried more than a 40% reduction in 3-yr mortality risk versus earlier ones, consistent with the authors' suggestion that many late in-breast events are new primaries rather than persistent disease.

For a radiation oncologist the most useful signal sits in the oLRR breakdown. Supraclavicular recurrences were the largest group, 81 of 165, and had the poorest course, 5-yr DDFS 12.1% against 31.5% after axillary recurrence, while oLRR incidence at 10 yrs rose from 4.4% with 1-3 nodes to 14.6% with ≥10. The source describes whole-breast irradiation only and gives no field, dose or nodal-coverage detail, so these cannot be classed as in-field or marginal failures. The dataset identifies the at-risk population; whether regional nodal irradiation changes its course was the question MA.20 and EORTC 22922 later took up with randomization, and this analysis cannot substitute for them.

Transferability is the main limit. These are 1984-1994 regimens: no taxanes, no HER2 testing or HER2-directed therapy, and premenopausal ER-positive women went without tamoxifen. Lumpectomy was not randomized, so the 2,669-pt subset reflects surgeon and pt selection. Margin-negative status was mandatory, which removes margin as an explanatory variable. Absolute locoregional failure rates and post-recurrence survival are both likely different with modern systemic therapy and salvage options. The paper leaves the treatment of an isolated recurrence explicitly open, noting a then-new IBCSG and NSABP trial of chemotherapy at relapse. Read it as a durable description of who fails locoregionally and how badly, not as a measure of what local therapy buys.