onc brain

The longer read

Before this trial, the boost question in DCIS was answered by extrapolation. Randomised evidence for a tumour bed boost came from invasive disease, principally the EORTC boost trial, and the DCIS trials that established whole breast irradiation after lumpectomy (NSABP B-17, EORTC 10853 and their contemporaries) never randomised the boost. Practice varied accordingly: some centres boosted DCIS by analogy with invasive cancer, others omitted it on the grounds that DCIS behaves differently and that extrapolating a boost benefit was unjustified. BIG 3-07/TROG 07.01 replaces the analogy with a direct answer, and the direction matches the invasive-disease experience: HR 0.47 for local recurrence, with 5-year freedom from local recurrence of 97.1% vs 92.7%.

The population matters for how far this transfers. Eligibility required non-low-risk DCIS and at least 1 mm of clear radial margin, so these are pts in whom a clinician was already planning WBI and in whom margin status was not the driver of recurrence risk. That makes the result cleanest for the patient most radiation oncologists see: adequately excised, higher-grade or otherwise non-low-risk DCIS. It says nothing about low-risk DCIS, where the live question is omission of RT altogether rather than escalation, and it does not speak to close or involved margins, where boost is often used as a substitute for re-excision.

The factorial structure is the design feature that gives the result reach. Boost was tested on both a conventional 50 Gy/25 fx and a moderately hypofractionated 42.5 Gy/16 fx backbone, so the benefit is not an artefact of a fractionation schedule many centres have abandoned. The abstract does not report the fractionation-sensitivity comparison itself, which means the trial cannot be cited from this source as evidence that hypofractionation is equivalent in DCIS; that part of the question has to be read from the full paper or later reports.

The skeptic's reading centres on two points. First, the trial was unmasked, and the toxicity endpoints that carry the cost side of the trade (G2+ breast pain 14% vs 10%, induration 14% vs 6%) are assessor-dependent; they could plausibly be over- or under-called. Local recurrence is more objective but surveillance intensity was not masked either. Second, local recurrence in DCIS is roughly half invasive and half in situ in most series, and the abstract reports no survival or breast cancer mortality signal. A halving of recurrence hazard from a low baseline, bought with a doubling of G2+ induration, is a trade that different pts will value differently, and the absolute gain will shrink in pts whose baseline risk sits at the low end of non-low-risk.

What would have to be true for this to be wrong? The effect would need to be driven by a subset, such as younger pts or particular grades, with no benefit elsewhere; the abstract gives no subgroup data to test that. Until those are available, the defensible read is that boost is now evidence-based for non-low-risk DCIS after WBI, and the decision becomes a shared-risk conversation about morbidity rather than a question of whether any benefit exists. The next open problem is how boost interacts with shorter WBI courses and with partial-breast approaches that this trial did not test.